© 2012-2013 Real-Time Innovations, Inc.

All rights reserved.

Printed in U.S.A. First printing.

December 2013.

Trademarks

Real-Time Innovations, RTI, DataBus, and Connext are trademarks or registered trademarks of Real-Time Innovations, Inc. All other trademarks used in this document are the property of their respective owners.

Copy and Use Restrictions

No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form (including electronic, mechanical, photocopy, and facsimile) without the prior written permission of Real- Time Innovations, Inc. The software described in this document is furnished under and subject to the RTI software license agreement. The software may be used or copied only under the terms of the license agreement.

Technical Support

Real-Time Innovations, Inc.

232 E. Java Drive

Sunnyvale, CA 94089

Phone:	(408) 990-7444
Email:	support@rti.com
Website:	https://support.rti.com/

ii

3.2	Building and Running a Request-Reply Example		............................................................................ 3-8
3.3	An Introduction to Connext .................................................................................................................		3-9
	3.3.1 An Overview of Connext Objects ..........................................................................................		3-9
	3.3.2	DomainParticipants .................................................................................................................	3-9
	3.3.3	Publishers and DataWriters ..................................................................................................	3-12
	3.3.4	Subscribers and DataReaders ...............................................................................................	3-14
	3.3.5	Topics........................................................................................................................................	3-17
	3.3.6	Keys and Samples...................................................................................................................	3-18
	3.3.7	Requesters and Repliers ........................................................................................................	3-18

4 Capabilities and Performance ........................................................		4-1
4.1	Automatic Application Discovery.......................................................................................................	4-1
	4.1.1 When to Set the Discovery Peers............................................................................................	4-2
	4.1.2 How to Set Your Discovery Peers ..........................................................................................	4-2
4.2	Customizing Behavior: QoS Configuration .......................................................................................	4-3
4.3	Compact, Type-Safe Data: Programming with Data Types .............................................................	4-5
	4.3.1 Using Built-in Types.................................................................................................................	4-8
	4.3.2 Using Types Defined at Compile Time..................................................................................	4-8
	4.3.3 Running with Dynamic Types..............................................................................................	4-14

5 Design Patterns for Rapid Development.........................................			5-1
5.1	Building and Running the Code Examples........................................................................................		5-2
5.2	Subscribing Only to Relevant Data .....................................................................................................		5-3
	5.2.1	Content-Based Filtering...........................................................................................................	5-4
	5.2.2 Lifespan and History Depth ...................................................................................................		5-5
	5.2.3	Time-Based Filtering ................................................................................................................	5-9
5.3	Accessing Historical Data when Joining the Network....................................................................		5-11
	5.3.1	Implementation........................................................................................................................	5-11
	5.3.2	Running & Verifying ..............................................................................................................	5-12
5.4	Caching Data Within the Middleware ..............................................................................................		5-12
	5.4.1	Implementation.......................................................................................................................	5-13
	5.4.2	Running & Verifying ..............................................................................................................	5-15
5.5	Receiving Notifications When Data Delivery Is Late .....................................................................		5-15
	5.5.1	Implementation.......................................................................................................................	5-16
	5.5.2	Running & Verifying ..............................................................................................................	5-18

6 Design Patterns for High Performance ............................................			6-1
6.1	Building and Running the Code Examples........................................................................................		6-1
6.2	Reliable Messaging ................................................................................................................................		6-3
	6.2.1	Implementation.........................................................................................................................	6-4
6.3	High Throughput for Streaming Data ................................................................................................		6-6
	6.3.1	Implementation.........................................................................................................................	6-8
	6.3.2	Running & Verifying ................................................................................................................	6-9
6.4	Streaming Data over Unreliable Network Connections.................................................................		6-10
	6.4.1	Implementation.......................................................................................................................	6-10

iv

Chapter 1 Welcome to RTI Connext!

RTI® Connext™ solutions provide a flexible data distribution infrastructure for integrating data sources of all types. At its core is the world's leading ultra-high performance, distributed net- working DataBus™. It connects data within applications as well as across devices, systems and networks. Connext also delivers large data sets with microsecond performance and granular quality-of-service control. Connext is a standards-based, open architecture that connects devices from deeply embedded real-time platforms to enterprise servers across a variety of networks. Connext provides:

❏Ultra-low latency, extremely-high throughput messaging

❏Industry-leading reliability and determinism

❏Connectivity for heterogeneous systems spanning thousands of applications

Connext is flexible; extensive quality-of-service (QoS) parameters adapt to your application, assuring that you meet your real-time, reliability, and resource usage requirements.

This chapter introduces basic concepts and summarizes how Connext addresses your high-per- formance needs. After this introduction, we'll jump right into building distributed systems. The rest of this guide covers:

❏First steps: Installing Connext and creating your first simple application.

❏Learning more: An overview of the APIs and programming model with a special focus on the communication model, data types and qualities of service.

❏Towards real-world applications: An introduction to meeting common real-world require- ments.

If you’d like to skip to the next chapter, click: Installing Connext.

1.1A Guide to the Provided Documentation

This document will introduce you to the power and flexibility of Connext. We invite you to explore further by referring to the wealth of available information, examples, and resources:

The Connext documentation includes:

❏Core Libraries and Utilities Getting Started Guide (RTI_CoreLibrariesAndUtilities_GettingStarted.pdf)—This document describes how to install the Connext Core Libraries and Utilities. It also lays out the benefits and con- cepts behind the product and takes you step-by-step through the creation of a simple example application. Developers should read this document first.

1-1

A Guide to the Provided Documentation

If you want to use the Connext Extensible Types feature, please read:

•Addendum for Extensible Types (RTI_CoreLibrariesAndUtilities_GettingStarted_ExtensibleTypesAddendum.pdf) Extensible Types allow you to define data types in a more flexible way. Your data types can evolve over time—without giving up portability, interoperability, or the expres- siveness of the DDS type system.

If you are using Connext on an embedded platform or with a database, you will find addi- tional documents that specifically address these configurations:

•Addendum for Embedded Systems (RTI_CoreLibrariesAndUtilities_GettingStarted_EmbeddedSystemsAddendum.pdf)

•Addendum for Database Setup (RTI_CoreLibrariesAndUtilities_GettingStarted_DatabaseAddendum.pdf).

❏What’s New

(RTI_CoreLibrariesAndUtilities_WhatsNew.pdf)—This document describes changes and enhancements in the current version of Connext. If you are upgrading from a previ- ous version, read this document first.

❏Release Notes and Platform Notes (RTI_CoreLibrariesAndUtilities_ReleaseNotes.pdf, RTI_CoreLibrariesAndUtilities_PlatformNotes.pdf)—These documents provide sys- tem requirements, compatibility, and other platform-specific information about the prod- uct, including specific information required to build your applications using Connext, such as compiler flags and libraries.

❏User’s Manual

(RTI_CoreLibrariesAndUtilities_UsersManual.pdf)—This document describes the fea- tures of the product and how to use them. It is organized around the structure of the Con- next APIs and certain common high-level tasks.

❏API Reference HTML Documentation

(ReadMe.html)—This extensively cross-referenced documentation, available for all sup- ported programming languages, is your in-depth reference to every operation and con- figuration parameter in the middleware. Even experienced Connext developers will often consult this information.

The Programming How To's provide a good place to begin learning the APIs. These are hyperlinked code snippets to the full API documentation. From the ReadMe.html file, select one of the supported programming languages, then scroll down to the Programming How To’s. Start by reviewing the Publication Example and Subscription Example, which provide step-by step examples of how to send and receive data with Connext.

Many readers will also want to look at additional documentation available online. In particular, RTI recommends the following:

❏Use the RTI Customer Portal (http://support.rti.com) to download RTI software, access documentation and contact RTI Support. The RTI Customer Portal requires a username and password. You will receive this in the email confirming your purchase. If you do not have this email, please contact license@rti.com. Resetting your login password can be done directly at the RTI Customer Portal.

❏The RTI Community website (http://community.rti.com) provides a wealth of knowl- edge to help you use RTI Connext DDS, including:

•Best Practices

•Example code for specific features, as well as more complete use-case examples,

•Solutions to common questions,

1-2

Why Choose Connext?

•A glossary,

•Downloads of experimental software,

•And more.

❏Whitepapers and other articles are available from http://www.rti.com/resources.

Of course, RTI also offers excellent technical support and professional services. To contact RTI Support, simply log into the Customer Portal, send email to support@rti.com, or call the tele- phone number provided for your region. We thank you for your consideration and wish you success in meeting your distributed challenge.

1.2Why Choose Connext?

Connext implements publish/subscribe networking for high-performance distributed applica- tions. It complies with the Data Distribution Service (DDS) standard from the Object Manage- ment Group (OMG). Developers familiar with JMS and other middleware will see similarities, but will also find that DDS is not just another MOM (message-oriented middleware) standard! Its unique peer-to-peer architecture and unprecedented flexibility delivers much higher perfor- mance and adaptation to challenging applications. DDS is the first truly real-time networking technology standard. Connext is by far the market leading implementation of DDS.

1.2.1Reduce Risk Through Performance and Availability

Connext provides top performance, whether measured in terms of latency, throughput, or real- time determinism1. One reason is its elegant peer-to-peer architecture.

Traditional messaging middleware requires dedicated servers to broker messages—introducing throughput and latency bottlenecks and timing variations. Brokers also increase system admin- istration costs and represent single points of failure within a distributed application, putting data reliability and availability at risk.

RTI doesn’t use brokers. Messages flow directly from publishers to subscribers with minimal overhead. All the functions of the broker, including discovery (finding data sources and sinks), routing (directing data flow), and naming (identifying data types and topics) are handled in a fully-distributed, reliable fashion behind the scenes. It requires no special server software or hardware.

Traditional message-oriented middleware implementations require a broker to forward every message, increasing latency and decreasing determinism and fault tolerance.

RTI's unique peer-to-peer architecture eliminates bottlenecks and single points of failure.

1. For up-to-date performance benchmarks, visit RTI on the web at http://www.rti.com/products/dds/bench- marks-cpp-linux.html.

1-3

Why Choose Connext?

The design also delivers high reliability and availability, with automatic failover, configurable retries, and support redundant publishers, subscribers, networks, and more. Publishers and subscribers can start in any order, and enter and leave the network at any time; the middleware will connect and disconnect them automatically. Connext provides fine-grained control over fail- ure behavior and recovery, as well as detailed status notifications to allow applications to react to situations such as missed delivery deadlines, dropped connections, and slow or unresponsive nodes.

The User's Manual has details on these and all other capabilities. This guide only provides an overview.

1.2.2Reduce Cost through Ease of Use and Simplified Deployment

Connext helps keep development and deployment costs low by:

❏Increasing developer productivity—Easy-to-use, standards-compliant DDS APIs get your code running quickly. DDS is the established middleware standard for real-time pub- lish/subscribe communication in the defense industry and is expanding rapidly in utili- ties, transportation, intelligence, finance, and other commercial industries.

❏Simplifying deployment—Connext automatically discovers connections, so you don't need to configure or manage server machines or processes. This translates into faster turn- around and lower overhead for your deployment and administration needs.

❏Reducing hardware costs—Traditional messaging products require dedicated servers or acceleration hardware in order to host brokers. The extreme efficiency and reduced over- head of RTI's implementation, on the other hand, allows you to achieve the same perfor- mance using standard off-the-shelf hardware, with fewer machines than the competition.

1.2.3Ensure Success with Unmatched Flexibility

Out of the box, RTI is configured to achieve simple data communications. However, when you need it, RTI provides a high degree of fine-grained, low-level control over the middleware, including:

❏The volume of meta-traffic sent to assure reliability.

❏The frequencies and time-outs associated with all events within the middleware.

❏The amount of memory consumed, including the policies under which additional mem- ory may be allocated by the middleware.

RTI’s unique and powerful Quality-of-Service (QoS) policies can be specified in configuration files so that they can be tested and validated independently of the application logic. When not specified, the middleware will use default values chosen to provide good performance for a wide range of applications.

The result is simple-to-use networking that can expand and adapt to challenging applications, both current and future. RTI eliminates what is perhaps the greatest risk of commercial middle- ware: outgrowing the capability or flexibility of the design.

1.2.4Connect Heterogeneous Systems

Connext provides complete functionality and industry-leading performance for a wide variety of programming languages and platforms, including:

❏C, C++, .Net1, Java, and Ada2 development platforms

1.RTI Connext .NET language binding is currently supported for C# and C++/CLI.

1-4

What Can Connext Do?

❏Windows, Linux, Solaris, AIX, and other enterprise-class systems

❏VxWorks, INTEGRITY, LynxOS, and other real-time and/or embedded systems

Applications written in different programming languages, running on different hardware under different operating systems, can interoperate seamlessly over Connext, allowing disparate appli- cations to work together in even very complex systems.

1.2.5Interoperate with Databases, Event Engines, and JMS Systems

RTI provides connections between its middleware core and many types of enterprise software. Simple-but-powerful integrations with databases, Complex Event Processing (CEP) engines, and other middlewares ensure that RTI can bind together your real-time and enterprise systems.

RTI also provides a Java Message Service (JMS) application programming interface. Connext

directly interoperates at the wire-protocol level with RTI Message Service, the world’s highest- performing JMS implementation1. RTI Message Service, with its standard JMS interface, then pro- vides integration with a wide range of enterprise service busses and application gateways.

For more information about interoperability with other middleware implementations, including IBM MQ Series, please consult your RTI account representative.

1.3What Can Connext Do?

Under the hood, Connext goes beyond basic publish-subscribe communication to target the needs of applications with high-performance, real-time, and/or low-overhead requirements. It features:

❏Peer-to-peer, publish-subscribe communications—The most elegant, flexible data com- munications model.

•Simplified distributed application programming

•Time-critical data flow with minimal latency

•Clear semantics for managing multiple sources of the same data.

•Customizable Quality of Service and error notification.

•Guaranteed periodic messages, with minimum and maximum rates set by sub- scriptions

•Notifications when applications fail to meet their deadlines.

•Synchronous or asynchronous message delivery to give applications control over the degree of concurrency.

•Ability to send the same message to multiple subscribers efficiently, including sup- port for reliable multicast with customizable levels of positive and negative mes- sage acknowledgement.

❏Request-Reply Communications (available with Connext Messaging)—As applications become more complex, it often becomes necessary to use other communication patterns in addition to publish-subscribe. Sometimes an application needs to get a one-time snap- shot of information; for example, to make a query into a database or retrieve configura- tion parameters that never change. Other times an application needs to ask a remote

2.Ada support requires a separate add-on product, Ada 2005 Language Support.

1.For more information about RTI Message Service, see http://www.rti.com/products/jms/index.html. For perfor- mance benchmark results, see http://www.rti.com/products/jms/latency-throughput-benchmarks.html.

1-5

What Can Connext Do?

application to perform an action on its behalf. To support these scenarios, Connext Mes- saging includes support for the request-reply communication pattern. The Requester (ser- vice consumer or client) sends a request message and waits for a reply message. The Replier (service provider) receives the request message and responds with a reply mes- sage.

❏Reliable messaging—Enables subscribing applications to customize the degree of reli- ability required. Reliability can be tuned; you can guarantee delivery no matter how many retries are needed or try messages only once for fast and deterministic perfor- mance. You can also specify any settings in between. No other middleware lets you make this critical trade off on a per-message stream basis.

❏Multiple communication networks—Multiple independent communication networks (domains), each using Connext, can be used over the same physical network to isolate unrelated systems or subsystems. Individual applications can participate in one or multi- ple domains.

❏Symmetric architecture—Makes your application robust. No central server or privileged nodes means your system is robust to application and/or node failures.

❏Dynamic—Topics, subscriptions, and publications can be added or removed from the system at any time.

❏Multiple network transports—Connext includes support for UDP/IP (IPv4 and IPv6)— including, for example, Ethernet, wireless, and Infiniband networks—and shared mem- ory transports. It also includes the ability to dynamically plug in additional network transports and route messages over them. It can be configured to operate over a variety of transport mechanisms, including backplanes, switched fabrics, and other networking technologies.

❏Multi-platform and heterogeneous system support—Applications based on Connext can communicate transparently with each other regardless of the underlying operating sys- tem or hardware. Consult the Release Notes to see which platforms are supported in this release.

❏Vendor neutrality and standards compliance—The Connext API complies with the DDS specification. On the network, it supports the open DDS Interoperability Protocol, Real- Time Publish Subscribe (RTPS), which is also an open standard from the OMG.

1-6

What Can Connext Do?

Am I Better Off Building My Own Middleware?

Sometimes application projects start with minimal networking needs. So it’s natural to consider whether building a simplified middleware in-house is a better alternative to purchasing a more- complex commercial middleware. While doing a complete Return on Investment (ROI) analysis is outside the scope of this document, with Connext, you get a rich set of high-performance net- working features by just turning on configuration parameters, often without writing a single line of additional code.

RTI has decades of experience with hundreds of applications. This effort created an integrated and time-tested architecture that seamlessly provides the flexibility you need to succeed now and grow into the future. Many features require fundamental support in the core of the middle- ware and cannot just be cobbled onto an existing framework. It would take many man-years of effort to duplicate even a small subset of the functionality to the same level of stability and reli- ability as delivered by RTI.

For example, some of the middleware functionality that your application can get by just enabling configuration parameters include:

❏Tuning reliable behavior for multicast, lossy and high-latency networks.

❏Supporting high data-availability by enabling redundant data sources (for example, “keep receiving data from source A. If source A stops publishing, then start receiving data from source B”), and providing notifications when applications enter or leave the network.

❏Optimizing network and system resources for transmission of periodic data by support- ing time-based filtering of data (example: “receive a sample no more than once per sec- ond”) and deadlines (example: “expect to receive a sample at least once per second”).

Writing network code to connect a few nodes is deceptively easy. Making that code scale, handle all the error scenarios you will eventually face, work across platforms, keep current with tech- nology, and adapt to unexpected future requirements is another matter entirely. The initial cost of a custom design may seem tempting, but beware; robust architectures take years to evolve. Going with a fire-tested, widely used design assures you the best performance and functionality that will scale with your distributed application as it matures. And it will minimize the pro- found cost of going down the wrong path.

1-7

What are the Connext Products?

1.4What are the Connext Products?

This Getting Started Guide describes RTI Connext DDS and Connext Messaging.

RTI Connext Professional Edition includes RTI Connext Messaging, RTI Connext Integrator, and RTI Connext Tools, plus an installer. See the RTI Connext Professional Edition Getting Started Guide for more information.

1.4.1RTI Connext DDS and Connext Messaging

Connext DDS addresses the sophisticated data bus requirements in complex systems including an API compliant with the Object Management Group (OMG) Data Distribution Service (DDS) specification. DDS is the leading data-centric publish/subscribe (DCPS) messaging standard for integrating distributed real-time applications. Connext DDS is the dominant industry implemen- tation with benefits including:

❏OMG-compliant DDS API

❏Advanced features to address complex systems

❏Advanced Quality of Service (QoS) support

❏Comprehensive platform and network transport support

❏Seamless interoperability with Connext Micro and Connext Messaging

Connext Messaging provides a versatile and highly scalable messaging middleware for develop- ing applications leveraging a variety of embedded and enterprise design patterns. Connext Mes- saging flexibility reduces development, integration and testing costs and enables rapid implementation of new system requirements. Benefits include:

❏Communication patterns for publish/subscribe, request-reply, command/response, point-to-point with queues or a hybrid of patterns for unique customization

❏Support for RTI Persistence Service for late joiners and RTI Recording Service for logging data for deep analysis and archiving

❏APIs for JMS and Connext DDS

❏Tools to monitor, analyze and debug your complete system

❏Seamless interoperability with Connext Micro and Connext DDS

This table lists the features provided with each product. The additional components provided with Connext Messaging are highlighted in blue.

Table 1.1 Comparing Connext DDS and Connext Messaging

√ = included; o = optional, † = legacy	Connext DDS	Connext Messaging
Core Libraries & Utilities
DDS - C, C++, C#, Java	√	√
DDS - Ada	o
Message Service (JMS API)		√
RTSJ Extension Kit	o	o

1-8

What are the Connext Products?

Table 1.1 Comparing Connext DDS and Connext Messaging

√ = included; o = optional, † = legacy	Connext DDS	Connext Messaging
Transports
Shared memory, UDPv4/6, TCP	√	√
TLS		√
Secure WAN (DTLS)		√
Limited-Bandwidth Plug-Ins	o	o
Core Capabilities
Publish-Subscribe Communication	√	√
Request-Reply Communication		√
Persistence Service		√
Enterprise Discovery	o	o
Tools
Interface compiler (rtiddsgen)	√	√
CORBA Compatibility Kit	o	o
Utilities (spy, ping)	√	√
Spreadsheet Add-in for Microsoft® Excel®	√	√
Wireshark	√	√
Launcher	√	√
Scope	†	†

1.4.2RTI Connext Integrator

RTI Connext Integrator is a flexible integration infrastructure for OT systems, legacy systems and bridging to IT applications while requiring little or no modification. Connext Integrator is based on a proven foundation of technology and tools used in highly diverse systems, often with extreme real-time performance requirements. Connext Integrator benefits include:

❏Support of a rich set of integration capabilities including Message Translator, Content Enricher, Content Remover, Splitter and Aggregator

❏Integration support for standards including JMS, SQL databases, file, socket, Excel, OPC, STANAG, LabVIEW, Web Services and more

❏Ability for users to create custom integration adapters

❏Database Service integration for Oracle, MySQL and other relational databases

❏Tools for visualizing, debugging and managing all systems in real-time

Connext Integrator includes these components:

❏RTI Routing Service

❏RTI Routing Service Adapter SDK

❏RTI Real-Time Connect

1-9

What are the Connext Products?

1.4.3RTI Connext Micro

RTI Connext Micro provides a small-footprint, modular messaging solution for resource-limited devices that have minimal memory and CPU power, and may not even be running an operating system. Additionally, Connext Micro is designed to become a certifiable component in high- assurance systems. RTI Connext Micro benefits include:

❏Accommodations for resource-constrained environments

❏Modular and user extensible architecture

❏Designed to be a certifiable component for safety-critical systems

❏Seamless interoperability with Connext DDS and Connext Messaging

1.4.4RTI Connext Tools

RTI Connext Tools includes a rich set of components to accelerate debugging and testing while easing management of deployed systems. These components include:

❏Administration Console

❏Analyzer

❏Distributed Logger

❏Monitor

❏Monitoring Library

❏Recording Service

1-10

Chapter 2 Installing Connext

Connext is delivered as static and dynamic libraries with development tools, examples, and doc- umentation that will help you use the libraries.

This chapter includes:

❏Installing Connext

❏License Management

❏Navigating the Directories

❏Selecting a Development Environment

Next Chapter —> Getting Started with Connext

2.1Installing Connext

Connext is available for installation in the following forms, depending on your needs and your license terms:

❏In a single package as part of RTI Connext Professional Edition, which includes both host

(development) and target (deployment) files. If you downloaded an evaluation of Con- next from the RTI website at no cost, this is the distribution you have. For installation instructions, please see the RTI Connext Professional Edition Getting Started Guide

(RTI_Connext_Professional_Edition_GettingStarted.pdf).

•For Windows users: Your installation package is an executable installer application that installs a specific Windows architecture.

•For Linux users: Your installation package is a .sh file that installs files for a specific Linux architecture.

❏In separate host and target packages. This distribution is useful for those customers who develop and deploy on different platforms, or who deploy on multiple platforms. RTI also delivers support for embedded platforms this way. To request access to additional RTI host or target platforms, please contact your RTI account representative.

•For Windows users: You have at least two .zip archives: one for your host platform and at least one target platform.

•For users of other operating systems: You have at least two .tar.gz archives: one for your host platform and at least one target platform.

2-1

Installing Connext

2.1.1Installing on a UNIX-Based System

If you are using the RTI Connext Professional Edition installer, please see the RTI Connext Profes- sional Edition Getting Started Guide for details.

If you are installing the RTI Core Libraries and Utilities independently (that is, you are not using the RTI Connext Professional Edition installer), the distribution is packaged in two or more .tar.gz files: a host archive that contains documentation, header files, and other files you need for devel- opment; and one or more target archives containing the libraries and other files you need for deployment. Unpack them as described below. You do not need to be logged in as root during installation.

In the following instructions, we assume your host distribution file is named

RTI_Connext_DDS_Host-5.x.y-i86Linux.tar.gz and your target file is named RTI_Connext_DDS_Target-5.x.y-.i86Linux2.6gcc4.4.5.tar.gz. Your filenames will be different depending on your architecture, whether you have RTI Connext DDS or RTI Connext Messaging, whether it is a license-managed version, and your release version number (in our example, the ‘x’ and ‘y’ represent release-specific numbers).

1.Make sure you have GNU’s version of the tar utility (which handles long file names). On Linux systems, this is the default tar executable. On Solaris systems, use gtar.

2.Create a directory for Connext. We will assume that you want to install under /opt/rti/ (you may replace references to /opt/rti/ with the directory of your choice).

3.Move the downloaded host and target files into your newly created directory.

4.Extract the host distribution first. For example:

gunzip RTI_Connext_DDS_Host-5.x.y-i86Linux.tar.gz gtar xvf RTI_Connext_DDS_Host-5.x.y-i86Linux.tar

5. Extract the target distribution(s)). For example:

gunzip RTI_Connext_DDS_Target-5.x.y-i86Linux.tar.gz

gtar xvf RTI_Connext_DDS_Host-5.x.y-i86Linux2.6gcc4.4.5.tar

Using our example path, you will end up with /opt/rti/ndds.5.x.y. (Note: If you have a license-managed distribution of Connext, the directory will be /opt/rti/ndds.5.x.y_lic.)

6.Optional—If you want to use the TCP transport included with Connext in a secure config- uration, you also need to install RTI TLS Support and OpenSSL. To purchase RTI TLS Sup- port, please contact your RTI account representative.

7.Read License Management (Section 2.2).

2.1.2Installing on a Windows System

If you are using the RTI Connext Professional Edition installer, please see the RTI Connext Profes- sional Edition Getting Started Guide for details.

If you are installing the RTI Core Libraries and Utilities independently (that is, you are not using the RTI Connext Professional Edition installer), unpack the .zip files as described below.

Depending on your needs, you will have two or more .zip archives (a host archive containing documentation, header files, and other files you need for development, and one or more target archives containing the libraries and other files you need for deployment). For example, you may have a host distribution file named RTI_Connext_DDS_Host-5.x.y-i86Win32/zip and a target file named RTI_Connext_DDS_Target-5.x.y-.i86Win32VS2010.zip. Your filenames will be different depending on your architecture, whether you have RTI Connext DDS or RTI Connext Messaging, whether it is a license-managed version, and your release version number (in our example, the ‘x’ and ‘y’ represent release-specific numbers).

2-2

License Management

Depending on your version of Windows and where you want to expand these files, your user account may or may not require administrator privileges.

1.Create a directory for Connext. We will assume that you want to install under C:\Pro- gram Files\RTI (you may replace references to C:\Program Files\RTI with the direc- tory of your choice).

2.Move the downloaded host and target files into your newly created directory.

3.Extract the distribution from the .zip files. You will need a zip file utility such as Win- Zip® to help you. Extract the host distribution first.

Using our example path, you will end up with C:\Program Files\RTI\ndds.5.x.y. (Note: If you have a license-managed distribution of Connext, the directory will be C:\Program Files\RTI\ndds.5.x.y_lic.)

4.Optional—If you want to use the TCP transport included with Connext in a secure config- uration, you also need to install RTI TLS Support and OpenSSL. To purchase RTI TLS Sup- port, please contact your RTI account representative.

5.Read License Management (Section 2.2).

2.2License Management

Some distributions of Connext require a license file in order to run (for example, those provided for evaluation purposes or on a subscription basis).

If your Connext distribution requires a license file, you will receive one from RTI via email.

If you have licenses files for both the RTI Core Libraries and Utilities and RTI Message Service, you can concatenate both into one file.

A single license file can be used to run on any architecture and is not node-locked. You are not required to run a license server.

2.2.1Installing the License File

Save the license file in any location of your choice; the locations checked by the middleware are listed below.

Each time your Connext application starts, it will look for the license file in the following loca- tions until it finds a valid license:

1.In your application’s current working directory, in a file called rti_license.dat.

2.In the location specified in the environment variable RTI_LICENSE_FILE, which you may set to point to the full path of the license file, including the filename (for example,

C:\RTI\my_license_file.dat).

3.In the Connext installation, in the file $NDDSHOME/rti_license.dat. (See Section 3.1.1.1 for details on NDDSHOME.)

As Connext attempts to locate and read your license file, you may (depending on the terms of the license) see a printed message with details about your license.

If the license file cannot be found or the license has expired, your application may be unable to initialize Connext, depending on the terms of the license. If that is the case, your application’s call to DomainParticipantFactory.create_participant() will return null, preventing communica- tion.

2-3

Navigating the Directories

If you have any problems with your license file, please email support@rti.com.

2.2.2Adding or Removing License Management

If you are using an RTI software distribution that requires a license file and your license file changes—for example, you receive a new license for a longer term than your original license— you do not need to reinstall Connext.

However, if you switch from a license-managed distribution of Connext to one of the same ver- sion that does not require license management, or visa versa, RTI recommends that you first uninstall your original distribution before installing your new distribution. Doing so will pre- vent you from inadvertently using a mixture of libraries from multiple installations.

2.3Navigating the Directories

Once you have installed Connext, you can access the example code used in this document under $(NDDSHOME}/example1. RTI supports the C, C++, C++/CLI, C#, Java, and Ada2 program- ming languages. While we will examine the C++, Java, and Ada examples in the following chap- ters, you can access similar code in the language of your choice.

The Connext directory tree is as follows:

$NDDSHOME	Root directory where Connext is installed
/class	Java library files
/doc	Documentation in HTML and PDF format
/example	Example code
/ADA2
/C
/CPP
/CPPCLI
/CSHARP
/JAVA
/QoS	Quality of Service (QoS) configuration files for the example code. These
	configuration files do not depend on the language bindings, and largely
	do not depend on the operating system either.
/include	Header files for C and C++ APIs, specification files for Ada
/jre	Java runtime environment files used by the rtiddsgen tool (you do not
	need to use this JRE to run your own Java applications)
/lib	Library files
/gnat	Ada project files
/GNATgcc	Ada libraries and library information files
/ReadMe.html	Start page for accessing the HTML documentation
/resource	Document format definitions and template files used by
	rtiddsgen
/scripts	RTI tools (this directory should be in your path)

1.See Section 3.1.1.1 for details on NDDSHOME.

2.Ada support requires a separate add-on product, Ada 2005 Language Support.

2-4

Selecting a Development Environment

The two most important directories for learning Connext are example and doc. The doc directory contains the Connext library information in PDF and HTML formats. You may want to book- mark the HTML directory since you will be referring to this page a lot as you explore the RTI technology platform.

Example code is shipped in C, C++, C#, Java, and Ada2. You can access the example code in a language of your choice from the respective locations under the example directory. See the instructions in each example’s README_ME.txt file. These examples include:

❏Hello_simple: This example demonstrates one of the simplest applications you can write with Connext: it does nothing but publish and subscribe to short strings of text. This example is described in detail in Chapter 3: Getting Started with Connext.

❏Hello_builtin, Hello_idl, Hello_dynamic1: These examples demonstrate more some of the unique capabilities of Connext: strongly typed data, QoS-based customization of behav- ior, and industry-leading performance. These examples are described in Chapter 4: Capabilities and Performance and Chapter 6: Design Patterns for High Performance.

❏helloWorldRequestReply: This example demonstrates how to use Requesters and Repli- ers (Section 3.3.7). The Replier is capable of computing the prime numbers below a certain positive integer; the Requester will request these numbers. The Replier provides the prime numbers as they are being calculated, sending multiple replies. See Building and Run- ning a Request-Reply Example (Section 3.2).

❏News: This example demonstrates a subset of the rich functionality Connext offers, including flexible support for historical and durable data, built-in data caching, powerful filtering capabilities, and tools for working with periodic data. This example is described in Chapter 5: Design Patterns for Rapid Development.

You can find more examples at http://www.rti.com/examples. This page contains example code snippets on how to use individual features, examples illustrating specific use cases, as well as performance test examples.

2.4Selecting a Development Environment

You can develop applications with Connext either by building and executing commands from a shell window, or by using a development environment like Microsoft® Visual Studio®, Eclipse™, or GPS from AdaCore2.

2.4.1Using the Command-Line Tools on UNIX-based Systems

For Java-based applications, you can use the following script files to build and execute:

❏example/JAVA/<example>/build.sh—Builds your Java source files; no parameters are needed.

❏example/JAVA/<example>/run.sh—Runs the main program in either Publisher or Sub- scriber mode. It accepts the following parameters:

•[pub|sub]—Run as publisher or subscriber

•–verbose—Increases output verbosity

(This script accepts other options, which we will discuss in later chapters.)

1.Hello_dynamic is not provided for Ada.

2.Ada support requires a separate add-on product, Ada 2005 Language Support.

2-5

Selecting a Development Environment

For C and C++ applications for UNIX-based systems, you can use the make command with this makefile:

❏example/[C|CPP]/<example>/make/Makefile.<architecture>

where <architecture> reflects the compiler, OS and processor combination for your devel- opment environment. If you do not see your architecture listed, see Generating Code with rtiddsgen (Section 4.3.2.1) for instructions on how to generate an example makefile.

For Ada applications2, use the make command with this makefile:

❏example/ADA/<example>/make/Makefile.<architecture>

where <architecture> reflects the compiler, OS and processor combination for your devel- opment environment. If you do not see your architecture listed, see Generating Code with rtiddsgen (Section 4.3.2.1) for instructions on how to generate an example makefile.

2.4.2Using the Command-Line Tools on Windows Systems

For Java-based applications, you can use the following script files to build and execute:

❏example\JAVA\<example>\build.cmd—Builds Java source files; no parameters are needed.

❏example\JAVA\<example>\run.cmd—Runs the main program, Hello, in either Pub- lisher or Subscriber mode. It accepts the following parameters:

•[pub|sub]—Run as publisher or subscriber

•–verbose—Increases output verbosity

(This script accepts other options, which we will discuss in later chapters.)

For Java users: The native libraries used by the RTI Java API require the Visual Studio 2005 ser- vice pack 1 redistributable libraries on the target machine. You can obtain this package from Microsoft or RTI.

For C, C++, and C# users: Please use Microsoft Visual Studio 2005, service pack 1 or later1, or Visual Studio 2008 to build and run the examples.

2.4.3Using Microsoft Visual Studio

Connext includes solutions and project files for Microsoft Visual Studio in exam- ple\[C|CPP|CSHARP]\<example>\win32.

To use these solution files:

1.Start Visual Studio.

2.Select File, Open, Project/Solution.

3.In the File dialog, select the solution file for your architecture; the solution file for Visual Studio 2005 for 32-bit platforms is example\[C|CPP|CSHARP]\<example>\win32\Hello-i86Win32VS2005.sln.

1.If you are using an earlier version of Visual Studio, you can obtain a no-cost edition of Visual Studio 2005 or 2008 from Microsoft's web site.

2-6

Chapter 3 Getting Started with Connext

This chapter gets you up and running with Connext. First, you will build and run your first Con- next-based application. Then we'll take a step back to learn about the general concepts in the middleware and show how they are applied in the example application you ran.

This chapter includes:

❏Building and Running “Hello, World” (Section 3.1)

❏Building and Running a Request-Reply Example (Section 3.2)

❏An Introduction to Connext (Section 3.3)

Next Chapter —> Chapter 4: Capabilities and Performance

3.1Building and Running “Hello, World”

Let’s start by compiling and running Hello World, a basic program that publishes information over the network.

For now, do not worry about understanding the code (we start covering it in Chapter 4: Capabil- ities and Performance). Use the following instructions to run your first middleware program using Connext.

3.1.1Step 1: Set Up the Environment

There are a few things to take care of before you start working with the example code.

3.1.1.1Set Up the Environment on Your Development Machine

a.Set the NDDSHOME environment variable.

Set the environment variable NDDSHOME to the Connext install directory. (Connext itself does not require that you set this environment variable. It is used in the scripts used to build and run the example code because it is a simple way to locate the install direc- tory. You may or may not choose to use the same mechanism when you create scripts to build and/or run your own applications.)

•On UNIX-based systems: This location may be /opt/rti/ndds.5.x.y.

•On Windows systems: This location may be C:\Program Files\RTI\ndds.5.x.y.

3-1

Building and Running “Hello, World”

If you have multiple versions of Connext installed:

As mentioned above, Connext does not require that you have the environment variable NDDSHOME set at run time. However, if it is set, the middleware will use it to load cer- tain configuration files. Additionally, you may have previously set your path based on the value of that variable. Therefore, if you have NDDSHOME set, be sure it is pointing to the right copy of Connext.

b.Update your path.

Add Connext's scripts directory to your path. This will allow you to run some of the sim- ple command-line utilities included in your distribution without typing the full path to the executable.

•On UNIX-based systems: Add the directory to your PATH environment variable.

•On Windows systems: Add the directory to your Path environment variable.

c.If you will be using the separate add-on product, Ada 2005 Language Support:

Add $NDDSHOME/lib/gnat to your ADA_PROJECT_PATH environment variable. This directory contains Ada project files that will be used in the generated example project file.

Make sure the Ada compiler, gprbuild, is in your path. The makefile used by the exam- ple assumes that gprbuild is in your path.

On UNIX-based systems: Add the path to gprbuild to your PATH environment variable.

d.Make sure Java is available.

If you plan to develop in a language other than Java, you do not need Java installed on your system and can safely skip this step.

If you will be going through the examples in Java, ensure that appropriate java and javac executables are on your path. They can be found within the bin directory of your JDK installation. The Release Notes list the Java versions that are supported.

On Linux systems: Note that GNU java (from the GNU Classpath project) is not sup- ported—and will typically not work—but is on the path by default on many Linux sys- tems.

e.Make sure the preprocessor is available.

Check whether the C preprocessor (e.g., cpp) is in your search path. This step is optional, but makes code generation with the rtiddsgen utility more convenient. Chapter 4: Capa- bilities and Performance describes how to use rtiddsgen.

On Windows systems:

•If you have Microsoft Visual Studio installed: Running the script vcvars32.bat, vsvars32.bat, or vcvarsall.bat (depending on your version of Visual Studio) will update the path for a given command prompt. If the Visual Studio installer did not add cl to your path already, you can run this script before running rtiddsgen.

•If you do not have Microsoft Visual Studio installed: This is often the case with Java users. You can either choose not to use a preprocessor or to obtain a no-cost version of Visual Studio from Microsoft's web site.

f.Get your project files ready.

If you installed Connext in a directory that is shared by multiple users (for example, in C:\Program Files on a Windows system, or /opt or /local on a UNIX-based system), you may prefer to build and edit the examples in a directory of your own so you do not inter- fere with other users. (Depending on your file permissions, you may have to copy them somewhere else.) If you would like, copy the directory ${NDDSHOME}/example to a location of your choice. Where you see ${NDDSHOME}/example mentioned in the instructions below, substitute your own copied directory instead.

3-2

Building and Running “Hello, World”

•On Windows systems: If you chose to install Connext in C:\Program Files or another system directory, Microsoft Visual Studio may present you with a warning message when you open a solution file from the installation directory. If you see this dialog, you may want to copy the example directory somewhere else, as described above.

•On UNIX-based systems: The makefiles that RTI provides with the example code are intended to be used with the GNU distribution of the make utility. On modern Linux systems, the make binary typically is GNU make. On other systems, GNU make is called gmake. For the sake of clarity, the name gmake is used below. Make sure that the GNU make binary is on your path before continuing.

3.1.1.2Set Up the Environment on Your Deployment Machine

Some configuration has to be done for the machine(s) on which you run your application; the RTI installer can’t do that for you, because those machines may or may not be the same as where you created and built the application.

a.Make sure Java is available.

If you are a Java user, see Step d in Section 3.1.1.1 for details.

b.Make sure the dynamic libraries are available.

Make sure that your application can load the Connext dynamic libraries. If you use C, C++, or Ada1 with static libraries (the default configuration in the examples covered in

this document), you can skip this step. However, if you plan to use dynamic libraries, or Java or .NET2 (which always use dynamic libraries), you will need to modify your envi- ronment as described here.

To see if dynamic libraries are supported for your machine’s architecture, see the RTI Core Libraries and Utilities Platform Notes3.

1.Ada support requires a separate add-on product, Ada 2005 Language Support.

2.RTI Connext .NET language binding is currently supported for C# and C++/CLI.

3.In the Platform Notes, see the “Building Instructions...” table for your target architecture.

3-3

Building and Running “Hello, World”

For more information about where the Windows OS looks for dynamic libraries, see: http://msdn.microsoft.com/en-us/library/ms682586.aspx.

•The dynamic libraries needed by C or C++ applications are in the directory ${NDDSHOME}/lib/<architecture>. The dynamic libraries needed by Ada1 applica- tions are in the directory ${NDDSHOME}/lib/GNATgcc/relocatable.

On UNIX-based systems: Add this directory to your LD_LIBRARY_PATH environ- ment variable.

On Mac OS systems: Add this directory to your DYLD_LIBRARY_PATH environ- ment variable.

On Windows systems: Add this directory to your Path environment variable or copy the DLLs inside into the directory containing your executable.

On AIX systems: Add this directory to your LIBPATH environment variable.

•The native dynamic libraries needed by Java applications are in the directory ${NDDSHOME}/lib/<architecture>. Your architecture name ends in jdk, e.g., i86Linux2.6gcc3.4.3jdk. (The gcc part—only present on UNIX-based architectures—

identifies the corresponding native architecture that relies on the same version of the C runtime library.) The native dynamic libraries needed by Ada1 applications are in the directory $NDDSHOME/lib/<architecture>. (The gcc part of <architecture> iden- tifies the corresponding native architecture that relies on the same version of the C run-time library.)

On UNIX-based systems: Add this directory to your LD_LIBRARY_PATH environ- ment variable.

On Mac OS systems: Add this directory to your DYLD_LIBRARY_PATH environ- ment variable.

On Windows systems: Add this directory to your Path environment variable.

On AIX systems: Add this directory to your LIBPATH environment variable.

•Java .jar files are in the directory ${NDDSHOME}/class. They will need to be on your application’s class path.

•On Windows systems: The dynamic libraries needed by .NET applications are in the directory %NDDSHOME%\lib\i86Win32dotnet2.0. You will need to either copy the

DLLs from that directory to the directory containing your executable, or add the directory containing the DLLs to your Path environment variable1. (If the .NET framework is unable to load the dynamic libraries at run time, it will throw a Sys- tem.IO.FileNotFoundException and your application may fail to launch.)

3.1.2Step 2: Compile the Hello World Program

The same example code is provided in C, C++, C#, Java, and Ada2. The following instructions cover C++, Java, and Ada in detail; the procedures for C and C# are very similar. The same source code can be built and run on different architectures.

The instructions also focus on Windows and UNIX-based systems. If you will be using an embedded platform, see the Embedded Systems Addendum

(RTI_Connext_GettingStarted_EmbeddedSystemsAddendum.pdf) for more instructions espe- cially for you.

1.The file nddsdotnet.dll (or nddsdotnetd.dll for debug) must be in the executable directory. Visual Studio will, by default, do this automatically.

2.Ada support requires a separate add-on product, Ada 2005 Language Support.

3-4

Building and Running “Hello, World”

C++ on Windows Systems: To build the example applications:

1.In the Windows Explorer, go to %NDDSHOME%\exam- ple\CPP\Hello_simple\win32 and open the Microsoft Visual Studio solution file for your architecture. For example, the file for Visual Studio 2005 for 32-bit platforms is

Hello-i86Win32VS2005.sln.

2.The Solution Configuration combo box in the toolbar indicates whether you are building debug or release executables; select Release. Select Build Solution from the Build menu.

C++ on UNIX-based Systems: To build the example applications:

1.From your command shell, go to ${NDDSHOME}/example/CPP/Hello_simple/.

2.Type:

>gmake -f make/Makefile.<architecture>

where <architecture> is one of the supported architectures; see the contents of the make directory for a list of available architectures. (If you do not see a makefile for your archi- tecture, please refer to Section 4.3.2.1 to learn how to generate a makefile or project files for your platform). This command will build a release executable. To build a debug ver- sion instead, type:

> gmake -f make/Makefile.<architecture> DEBUG=1

Java on Windows Systems: To build the example applications:

1.From your command shell, go to %NDDSHOME%\example\JAVA\ Hello_simple\.

2.Type:

>build

Java on UNIX-

based Systems: To build the example applications:

1. From your command shell, go to ${NDDSHOME}/example/JAVA/Hello_simple.

2. Type:	> ./build.sh

3-5

Building and Running “Hello, World”

Ada on UNIX-based Systems1: To build the example applications:

1. From your command shell, go to ${NDDSHOME}/example/ADA/Hello_simple/.

2. Type:	> gmake -f make/Makefile.<architecture>
where <architec-

ture> is one of the supported architectures; see the contents of the make directory for a list of available architectures. (If you do not see a makefile for your architecture, please refer to Section 4.3.2.1 to learn how to generate a makefile or project files for your plat- form). This command will build a release executable. To build a debug version instead,

>gmake -f make/Makefile.<architecture> DEBUG=1

3.1.3Step 3: Start the Subscriber

C++: To start the subscribing application:

❏On a Windows system: From your command shell, go to

%NDDSHOME%\example\CPP\Hello_simple and type:

>objs\<architecture>\HelloSubscriber.exe

where <architecture> is one of the supported architectures; see the contents of the win32 directory for a list of available architectures. For example, the Windows architecture name corresponding to 32-bit Visual Studio 2005 is i86Win32VS2005.

❏On a UNIX-based system: From your command shell, go to ${NDDSHOME}/example/CPP/Hello_simple and type:

>objs/<architecture>/HelloSubscriber

where <architecture> is one of the supported architectures; see the contents of the make directory for a list of available architectures. For example, the architecture name corre- sponding to Red Hat Enterprise Linux 5 is i86Linux2.6gcc4.1.1.

Java: To start the subscribing application:

(As described above, you should have already set your path appropriately so that the example application can load the native libraries on which Connext depends. If you have not, you can set the variable RTI_EXAMPLE_ARCH in your command shell—e.g., to i86Win32jdk or i86Linux2.6gcc4.1.1.jdk—and the example launch scripts will use it to set your path for you.)

❏On a Windows system: From your command shell, go to

%NDDSHOME%\example\JAVA\Hello_simple and type:

>runSub

❏On a UNIX-based system: From your command shell, go to ${NDDSHOME}/example/JAVA/Hello_simple and type:

>./runSub.sh

1. Ada support requires a separate add-on product, Ada 2005 Language Support.

3-6

Building and Running “Hello, World”

Ada1: To start the subscribing application:

❏On a UNIX-based system: From your command shell, go to ${NDDSHOME}/example/ADA/Hello_simple and type:

>objs/<architecture>/hellosubscriber

Where <architecture> is one of the supported architectures; see the content of the make directory for a list of available architectures. For example, the architecture name corre- sponding to Red Hat Enterprise Linux 5.5 is i86Linux2.6gcc4.1.2.

3.1.4Step 4: Start the Publisher

Connext interoperates across all of the programming languages it supports, so you can choose whether to run the publisher in the same language you chose for the subscriber or a different language.

C++: To start the publishing application:

❏On a Windows system: From a different command shell, go to

%NDDSHOME%\example\CPP\Hello_simple and type:

>objs\<architecture>\HelloPublisher.exe

where <architecture> is one of the supported architectures; see the contents of the win32 directory for a list of available architectures. For example, the Windows architecture name corresponding to 32-bit Visual Studio 2005 is i86Win32VS2005.

❏On a UNIX-based system: From a different command shell, go to ${NDDSHOME}/example/CPP/Hello_simple and type:

>objs/<architecture>/HelloPublisher

where <architecture> is one of the supported architectures; see the contents of the make directory for a list of available architectures. For example, the architecture name corre- sponding to Red Hat Enterprise Linux 5 is i86Linux2.6gcc4.1.1.

Java: To start the publishing application:

(As described above, you should have already set your path appropriately so that the example application can load the native libraries on which Connext depends. If you have not, you can set the variable RTI_EXAMPLE_ARCH in your command shell—e.g., to i86Win32jdk or i86Linux2.6gcc4.1.1.jdk—and the example launch scripts will use it to set your path for you.)

❏On a Windows system: From a different command shell, go to

%NDDSHOME%\example\JAVA\Hello_simple and type:

>runPub

❏On a UNIX-based system: From a different command shell, go to ${NDDSHOME}/example/JAVA/Hello_simple and type:

>./runPub.sh

Ada2: To start the publishing application:

❏On a UNIX-based system: From a different command shell, go to /example/ADA/ Hello_simple and type:

>objs/<architecture>/hellopublisher

1.Ada support requires a separate add-on product, Ada 2005 Language Support.

3-7

Building and Running a Request-Reply Example

Where <architecture> is one of the supported architectures; see the contents of the make directory for a list of available architectures. For example, the architecture name corre- sponding to Red Hat Enterprise Linux 5.5 is i86Linux2.6gcc4.1.2.

If you typed "Hello, World" in the publishing application, you should see output similar to the following:

Congratulations! You’ve run your first Connext program!

3.2Building and Running a Request-Reply Example

Important! This section describes the Request-Reply communication pattern, which is only available with RTI Connext Messaging. It is not provided with RTI Connext DDS.

Requesters and Repliers (Section 3.3.7) contains an introduction to the Connext Request-Reply API. More detailed information is available in the RTI Core Libraries and Utilities User's Manual (see Part 4: Request-Reply Communication) and in the API Reference HTML documentation (open NDDSHOME/ReadMe.html and select a programming language; then select Modules, RTI Connext Request-Reply API Reference). See A Guide to the Provided Documentation (Section 1.1).

RTI Connext Messaging provides the libraries that you will need (in addition to the DDS libraries) when compiling an application that uses the Request-Reply API.

❏In C, you need the additional rticonnextmsgc libraries and to use a set of macros that instantiate type-specific code. You will see how to do this in the code example.

❏In C++, you need the additional rticonnextmsgcpp libraries and the header file ndds/ ndds_requestreply_cpp.h.

❏In Java, you need an additional JAR file: rticonnextmsg.jar.

❏In .NET (C# and C++/CLI), you need the additional assembly rticonnextmsgdotnet.dll (for .NET 2.0) or rticonnextmsgdotnet40.dll (for .NET 4.0).

To set up your environment follow the same instructions in Section 3.1.1.

The Request-Reply examples are located under your installation directory (NDDSHOME):

2. Ada support requires a separate add-on product, Ada 2005 Language Support.

3-8

An Introduction to Connext

❏example/C/helloWorldRequestReply

❏example/CPP/ helloWorldRequestReply

❏example/CSHARP/helloWorldRequestReply

❏example/JAVA/helloWorldRequestReply

To compile the examples, follow the instructions in Section 3.1.2. Similar makefiles for UNIX- based systems and scripts for Java and Visual Studio projects are provided in these examples. Instructions for running the examples are in READ_ME.txt in the example directories. See the instructions in each example’s README_ME.txt file.

3.3An Introduction to Connext

Connext is network middleware for real-time distributed applications. It provides the communi- cations service that programmers need to distribute time-critical data between embedded and/ or enterprise devices or nodes. Connext uses a publish-subscribe communications model to make data-distribution efficient and robust. Connext Messaging also supports for the Request-Reply communication pattern.

Connext implements the Data-Centric Publish-Subscribe (DCPS) API of the OMG’s specification, Data Distribution Service (DDS) for Real-Time Systems. DDS is the first standard developed for the needs of real-time systems, providing an efficient way to transfer data in a distributed system. With Connext, you begin your development with a fault-tolerant and flexible communications infrastructure that will work over a wide variety of computer hardware, operating systems, lan- guages, and networking transport protocols. Connext is highly configurable, so programmers can adapt it to meet an application’s specific communication requirements.

3.3.1An Overview of Connext Objects

The primary objects in Connext are:

❏DomainParticipants

❏Publishers and DataWriters

❏Subscribers and DataReaders

❏Topics

❏Keys and Samples

❏Requesters and Repliers

3.3.2DomainParticipants

A domain is a concept used to bind individual applications together for communication. To com- municate with each other, DataWriters and DataReaders must have the same Topic of the same data type and be members of the same domain. Each domain has a unique integer domain ID.

Applications in one domain cannot subscribe to data published in a different domain. Multiple domains allow you to have multiple virtual distributed systems on the same physical network. This can be very useful if you want to run multiple independent tests of the same applications. You can run them at the same time on the same network as long as each test runs in a different domain. Another typical configuration is to isolate your development team from your test and production teams: you can assign each team or even each developer a unique domain.

3-9

An Introduction to Connext

Figure 3.1 Connext Components

DomainParticipants are used to create Topics, Publishers, DataWriters, Subscribers, and DataReaders in the corresponding domain.

Figure 3.2 Segregating Applications with Domains

The following code shows how to instantiate a DomainParticipant. You can find more informa- tion about all of the APIs in the API Reference HTML documentation.

3-10

An Introduction to Connext

To create a DomainParticipant in C++:

participant = DDSDomainParticipantFactory::get_instance()->

create_participant(
0,	/*	Domain	ID */
DDS_PARTICIPANT_QOS_DEFAULT, /*		QoS */
NULL,	/*	Listener */
DDS_STATUS_MASK_NONE);

Here is the same logic in Java:

participant = DomainParticipantFactory.get_instance().create_participant(

0, // Domain ID DomainParticipantFactory.PARTICIPANT_QOS_DEFAULT, null, // Listener

StatusKind.STATUS_MASK_NONE);

In Ada1, the code to create a DomainParticipant looks like this:

participant := DDS.DomainParticipantFactory.Get_Instance.Create_Participant (0, -- Domain ID

DDS.DomainParticipantFactory.PARTICIPANT_QOS_DEFAULT, -- QoS null, -- Listener

DDS.STATUS_MASK_NONE);

As you can see, there are four pieces of information you supply when creating a new DomainPar- ticipant:

❏The ID of the domain to which it belongs.

❏Its qualities of service (QoS). The discussion on Page 3-12 gives a brief introduction to the concept of QoS; you will learn more in Chapter 4: Capabilities and Performance.

❏Its listener and listener mask, which indicate the events that will generate callbacks to the DomainParticipant. You will see a brief example of a listener callback when we discuss DataReaders below. For a more comprehensive discussion of the Connext status and notifi- cation system, see Chapter 4 in the RTI Core Libraries and Utilities User’s Manual.

1. Ada support requires a separate add-on product, Ada 2005 Language Support.

3-11

An Introduction to Connext

What is QoS?

Fine control over Quality of Service (QoS) is perhaps the most important feature of Connext. Each data producer-consumer pair can establish independent quality of service (QoS) agreements—even in many-to- many topologies. This allows applications to support extremely complex and flexible data-flow requirements.

QoS policies control virtually every aspect of Connext and the underlying communications mechanisms. Many QoS policies are implemented as "contracts" between data producers (DataWriters) and consumers (DataReaders); producers offer and consumers request levels of service. The middleware is responsible for determining if the offer can satisfy the request, thereby establishing the communication or indicating an incompatibility error. Ensuring that participants meet the level-of-service contracts guarantees predictable operation. For example:

❏Periodic producers can indicate the speed at which they can publish by offering guaranteed update deadlines. By setting a deadline, a producer promises to send updates at a minimum rate. Consumers may then request data at that or any slower rate. If a consumer requests a higher data rate than the producer offers, the middleware will flag that pair as incompatible and notify both the publishing and subscribing applications.

❏Producers may offer different levels of reliability, characterized in part by the number of past data samples they store for retransmission. Consumers may then request differing levels of reliable delivery, ranging from fast-but-unreliable "best effort" to highly reliable in-order delivery. This provides per- data-stream reliability control. A single producer can support consumers with different levels of reliability simultaneously.

Other QoS policies control when the middleware detects nodes that have failed, set delivery order, attach user data, prioritize messages, set resource utilization limits, partition the system into namespaces, control durability (for fault tolerance) and much more. The Connext QoS policies offer unprecedented flexible communications control. The RTI Core Libraries and Utilities User's Manual contains details about all available QoS policies.

3-12

An Introduction to Connext

Figure 3.3 Entities Associated with Publications

The following code shows how to create a DataWriter in C++:

data_writer = participant->create_datawriter( topic,

DDS_DATAWRITER_QOS_DEFAULT, /* QoS */ NULL, /* Listener */ DDS_STATUS_MASK_NONE);

In Ada1 it looks like this:

data_writer := participant.Create_DataWriter (topic,

DDS.Publisher.DATAWRITER_QOS_DEFAULT, -- QoS null, -- Listener

DDS.STATUS_MASK_NONE);

As you can see, each DataWriter is tied to a single topic. All data published by that DataWriter will be directed to that Topic.

As you will learn in the next chapter, each Topic—and therefore all DataWriters for that Topic—is associated with a particular concrete data type. The write operation, which publishes data, is type safe, which means that before you can write data, you must perform a type cast.

The code looks like this in C++:

string_writer = DDSStringDataWriter::narrow(data_writer);

1. Ada support requires a separate add-on product, Ada 2005 Language Support.

3-13

An Introduction to Connext

In Java:

StringDataWriter dataWriter =

(StringDataWriter) participant.create_datawriter( topic,

Publisher.DATAWRITER_QOS_DEFAULT, null, // listener StatusKind.STATUS_MASK_NONE);

In Ada1:

string_writer := DDS.Builtin_String_DataWriter.Narrow(data_writer);

Note that in this particular code example, you will not find any reference to the Publisher class. In fact, creating the Publisher object explicitly is optional, because many applications do not have the need to customize any behavior at that level. If you choose not to create a Publisher, the mid- dleware will implicitly choose an internal Publisher object. If you do want to create a Publisher explicitly, create it with a call to participant.create_publisher() (you can find more about this method in the API Reference HTML documentation) and then simply replace the call to partici- pant.create_datawriter() with a call to publisher.create_datawriter().

3-14

An Introduction to Connext

The following code demonstrates how to create a DataReader:

data_reader = participant->create_datareader( topic,

DDS_DATAREADER_QOS_DEFAULT, /* QoS */

&listener,

/* Listener */

DDS_DATA_AVAILABLE_STATUS);

Each DataReader is tied to a single topic. A DataReader will only receive data that was published on its Topic.

Here is the analogous code in Java:

StringDataReader dataReader =

(StringDataReader) participant.create_datareader(

topic,
Subscriber.DATAREADER_QOS_DEFAULT, //		QoS
new HelloSubscriber(),	//	Listener
StatusKind.DATA_AVAILABLE_STATUS);

And in Ada1:

dataReader := DDS. Builtin_String_DataReader.Narrow (participant.Create_DataReader

(topic.As_TopicDescription, DDS.Subscriber.DATAREADER_QOS_DEFAULT, -- QoS readerListener'Unchecked_Access, -- Listener DDS.DATA_AVAILABLE_STATUS));

Connext provides multiple ways for you to access your data: you can receive it asyn- chronously in a listener, you can block your own thread waiting for it to arrive using a helper object called a WaitSet, or you can poll in a non-blocking fashion. This example uses the former mechanism, and you will see that it passes a non-NULL listener to the create_datareader() method. The listener mask (DATA_AVAILABLE_STATUS) indi- cates that the application is only interested in receiving notifications of newly arriving data.

1. Ada support requires a separate add-on product, Ada 2005 Language Support.

3-15

An Introduction to Connext

Let’s look at the callback implementation in C++:

retcode = string_reader->take_next_sample( ptr_sample,

info);

if (retcode == DDS_RETCODE_NO_DATA) { /* No more samples */

break;

} else if (retcode != DDS_RETCODE_OK) {

cerr << "Unable to take data from data reader, error " << retcode << endl;

return;

}

And in Ada1:

loop

begin

data_reader.Read_Next_Sample (ptr_sample, sample_info'Access); if sample_info.Valid_Data then

Ada.Text_IO.Put_Line (DDS.To_Standard_String (ptr_sample)); end if;

exception

when DDS.NO_DATA =>

-- No more samples exit;

when others => Self.receiving := FALSE; exit;

end; end loop;

The take_next_sample() method retrieves a single data sample (i.e., a message) from the DataReader, one at a time without blocking. If it was able to retrieve a sample, it will return DDS_RETCODE_OK. If there was no data to take, it will return DDS_RETCODE_NO_DATA. Finally, if it tried to take data but failed to do so because it encountered a problem, it will return DDS_RETCODE_ERROR or another DDS_ReturnCode_t value (see the API Reference HTML documentation for a full list of error codes).

Connext can publish not only actual data to a Topic, but also meta-data indicating, for example, that an object about which the DataReader has been receiving data no longer exists. In the latter case, the info argument to take_next_sample() will have its valid_data flag set to false.

1. Ada support requires a separate add-on product, Ada 2005 Language Support.

3-16

An Introduction to Connext

This simple example is interested only in data samples, not meta-data, so it only processes “valid” data:

if (info.valid_data) {

// Valid (this isn't just a lifecycle sample): print it cout << ptr_sample << endl;

}

Let’s see the same thing in Java:

try {

String sample = stringReader.take_next_sample(info); if (info.valid_data) {

System.out.println(sample);

}

} catch (RETCODE_NO_DATA noData) {

//No more data to read break;

}catch (RETCODE_ERROR e) {

//An error occurred e.printStackTrace();

}

And in Ada1:

if sample_info.Valid_Data then

--Valid, print it

Ada.Text_IO.Put_Line (DDS.To_Standard_String (ptr_sample)); end if;

Note that in this particular code example, you will not find any reference to the Subscriber class. In fact, as with Publishers, creating the Subscriber object explicitly is optional, because many applications do not have the need to customize any behavior at that level. If you, like this exam- ple, choose not to create a Subscriber, the middleware will implicitly choose an internal Subscriber object. If you do want to create a Subscriber explicitly, create it with a call to partici- pant.create_subscriber (you can find more about this method in the API Reference HTML docu- mentation) and then simply replace the call to participant.create_datareader with a call to subscriber.create_datareader.

3.3.5Topics

Topics provide the basic connection points between DataWriters and DataReaders. To communi- cate, the Topic of a DataWriter on one node must match the Topic of a DataReader on any other node.

A Topic is comprised of a name and a type. The name is a string that uniquely identifies the Topic within a domain. The type is the structural definition of the data contained within the Topic; this capability is described in Chapter 4: Capabilities and Performance.

You can create a Topic with the following C++ code:

	topic = participant->create_topic(
	"Hello, World",		/* Topic	name*/
	DDSStringTypeSupport::get_type_name(), /* Type			name */
	DDS_TOPIC_QOS_DEFAULT,		/* Topic	QoS */
	NULL,		/* Listener */
	DDS_STATUS_MASK_NONE);

1. Ada support requires a separate add-on product, Ada 2005 Language Support.

3-17

An Introduction to Connext

Besides the new Topic’s name and type, an application specifies three things:

❏A suggested set of QoS for DataReaders and DataWriters for this Topic.

❏A listener and listener mask that indicate which events the application wishes to be noti- fied of, if any.

In this case, the Topic’s type is a simple string, a type that is built into the middleware. Let’s see the same logic in Java:

Topic topic = participant.create_topic(
"Hello, World",	// Topic name
StringTypeSupport.get_type_name(),	// Type name
DomainParticipant.TOPIC_QOS_DEFAULT, // QoS
null,	// Listener
StatusKind.STATUS_MASK_NONE);

In Ada1 it looks like this:

topic := participant.Create_Topic
(DDS.To_DDS_Builtin_String ("Hello, World"), --	Topic name
DDSStringTypeSupport.Get_Type_Name, -- Type	name
DDS.DomainParticipant.TOPIC_QOS_DEFAULT, --	Topic QoS
null, -- Listener
DDS.STATUS_MASK_NONE);

3.3.6	Keys and Samples
	The data values associated with a Topic can change over time. The
		A sample is analogous to a
	different values of the Topic passed between applications are called
		message in other publish-
	samples.
		subscribe middleware.
	An application may use a single Topic to carry data about many
	objects. For example, a stock-trading application may have a single topic, "Stock Price," that it
	uses to communicate information about Apple, Google, Microsoft, and many other companies.
	Similarly, a radar track management application may have a single topic, "Aircraft Position,"
	that carries data about many different airplanes and other vehicles. These objects within a Topic
	are called instances. For a specific data type, you can select one or more fields within the data
	type to form a key. A key is used to uniquely identify one instance of a Topic from another
	instance of the same Topic, very much like how the primary key in a database table identifies one
	record or another. Samples of different instances have different values for the key. Samples of the
	same instance of a Topic have the same key. Note that not all Topics have keys. For Topics without
	keys, there is only a single instance of that Topic.

3.3.7Requesters and Repliers

Important! This section describes the Request-Reply communication pattern, which is only available with RTI Connext Messaging.

Requesters and Repliers provide a way to use the Request-Reply communication pattern on top of the previously described Connext entities.

An application uses a Requester to send requests to a Replier; another application using a Replier receives a request and can send one or more replies for that request. The Requester that sent the request (and only that one) will receive the reply (or replies).

1.Ada support requires a separate add-on product, Ada 2005 Language Support.

3-18

An Introduction to Connext

A Requester uses an existing DomainParticipant to communicate through a domain. It owns a DataWriter for writing requests and a DataReader for receiving replies.

Similarly, a Replier uses an existing DomainParticipant to communicate through a domain and owns a DataReader for receiving requests and a DataWriter for writing replies.

Figure 3.5 Request-Reply Overview

The Reply Topic filters samples so replies are received by exactly one Requester—the one that wrote the related request sample.

You can specify the QoS for the DataWriters and DataReaders that Requesters and Repliers create.

The following code shows how to create a Requester in C++:

Requester<Foo, Bar> * requester =

new Requester<Foo, Bar>(participant, "MyService");

In Java it looks like this:

Requester<Foo, Bar> requester = new Requester<Foo, Bar>( participant, "MyService",

FooTypeSupport.get_instance(),

BarTypeSupport.get_instance());

As you can see, we are passing an existing DomainParticipant to the constructor.

Foo is the request type and Bar is the reply type. In Compact, Type-Safe Data: Programming with Data Types (Section 4.3), you will learn what types you can use and how to create them.

The constructor also receives a string "MyService." This is the service name, and is used to create the Request Topic and the Reply Topic. In this example, the Requester will create a Request Topic called "MyServiceRequest" and a Reply Topic called "MyServiceReply."

Creating a Replier is very similar. The following code shows how to create a Replier in C++:

Replier<Foo, Bar> * replier = new Replier<Foo, Bar>(

participant, "TestService");

3-19

An Introduction to Connext

And in Java:

Replier<Foo, Bar> replier = new Replier<Foo, Bar>(

participant, "TestService", FooTypeSupport.get_instance(), BarTypeSupport.get_instance());

This Replier will communicate with the Requester we created before, because they use the same service name (hence the topics are the same) and they use compatible QoS (the default).

More example code is available for C++, C, Java and C# as part of the API Reference HTML doc- umentation, under Modules, Programming How-To’s, Request-Reply Examples.

3-20

Chapter 4 Capabilities and Performance

In the previous chapter, you learned the basic concepts in Connext and applied them to a simple "Hello, World" application. In this chapter, you will learn more about some of the powerful and unique benefits of Connext:

❏A rich set of functionality, implemented for you by the middleware so that you don't have to build it into your application. Most of this functionality—including sophisticated data filtering and expiration, support for durable historical data, and built-in support for peri- odic data and deadline enforcement—can be defined partially or even completely in declarative quality-of-service (QoS) policies specified in an XML file, allowing you to examine and update your application's configuration without rebuilding or redeploying it. See Customizing Behavior: QoS Configuration (Section 4.2) for more information about how to configure QoS policies. Chapter 5: Design Patterns for Rapid Development describes how to reduce, filter, and cache data as well as other common functional design patterns.

❏Compact, type-safe data. The unique and expressive data-typing system built into Con- next supports not only opaque payloads but also highly structured data. It provides both static and dynamic type safety—without the network overhead of the "self-describing" messages of other networking middleware implementations. See Compact, Type-Safe Data: Programming with Data Types (Section 4.3) for more information.

❏Industry-leading performance. Connext provides industry-leading latency, throughput, and jitter performance. Chapter 6: Design Patterns for High Performance provides spe- cific QoS configuration examples to help you get started. You can quickly see the effects of the changes you make using the code examples described in that chapter. You can benchmark the performance of Connext on your own systems with the RTI Example Per- formance Test. You can download the Example Performance Test from http:// www.rti.com/examples/.

❏You can also review the data from several performance benchmarks here: http:// www.rti.com/products/dds/benchmarks-cpp-linux.html.

Next Chapter —> Chapter 5: Design Patterns for Rapid Development

4.1Automatic Application Discovery

As you’ve been running the code example described in this guide, you may have noticed that you have not had to start any server processes or configure any network addresses. Its built-in automatic discovery capability is one important way in which Connext differs from other net-

4-1

Automatic Application Discovery

working middleware implementations. It is designed to be low-overhead and require minimal configuration, so in many cases there is nothing you need to do; it will just work. Nevertheless, it’s helpful to understand the basics so that you can decide if and when a custom configuration is necessary.

Before applications can communicate, they need to “discover” each other. By default, Connext applications discover each other using shared memory or UDP loopback if they are on the same host or using multicast1 if they are on different hosts. Therefore, to run an application on two or more computers using multicast, or on a single computer with a network connection, no changes are needed. They will discover each other automatically! The chapter on Discovery in the RTI Core Libraries and Utilities User’s Manual describes the process in more detail.

If you want to use computers that do not support multicast (or you need to use unicast for some other reason), or if you want to run on a single computer that does not have a network connec- tion (in which case your operating system may have disabled your network stack), there is a simple way to control the discovery process—you won’t even have to recompile. Application discovery can be configured through the NDDS_DISCOVERY_PEERS environment variable or in your QoS configuration file.

4.1.1When to Set the Discovery Peers

There are only a few situations in which you must set the discovery peers:

(In the following, replace N with the number of Connext applications you want to run.)

1.If you cannot use multicast2:

Set your discovery peers to a list of all of the hosts that need to discover each other. The list can contain hostnames and/or IP addresses; each entry should be of the form

N@builtin.udpv4://<hostname|IP>.

2.If you do not have a network connection:

Some operating systems—for example, Microsoft Windows—disable some functionality of their network stack when they detect that no network interfaces are available. This can cause problems when applications try to open network connections.

•If your system supports shared memory2, set your discovery peers to N@buil- tin.shmem://. This will enable the shared memory transport only.

•If your system does not support shared memory (or it is disabled), set your discovery peers to the loopback address, N@builtin.udpv4://127.0.0.1.

4.1.2How to Set Your Discovery Peers

As stated above, in most cases you do not need to set your discovery peers explicitly. If setting them is required, there are two easy ways to do so:

❏Set the NDDS_DISCOVERY_PEERS environment variable to a comma-separated list of the names or addresses of all the hosts that need to discover each other.

• On Windows systems: For example:

set NDDS_DISCOVERY_PEERS=3@builtin.udpv4://mypeerhost1,\

4@builtin.udpv4://mypeerhost2

1.With the exception of LynxOS. On LynxOS systems, multicast is not used for discovery by default unless NDDS_DISCOVERY_PEERS is set.

2.To see if your platform supports RTI’s multicast network and shared memory transports, see the RTI Core Libraries and Utilities Platform Notes (RTI_Connext_PlatformNotes.pdf).

4-2

Customizing Behavior: QoS Configuration

•On UNIX-based systems: For example, if you are using csh or tcsh: setenv NDDS_DISCOVERY_PEERS 3@builtin.udpv4://mypeerhost1,\ 4@builtin.udpv4://mypeerhost2

❏Set the discovery peer list in your XML QoS configuration file.

For example, to turn on shared memory only:

<participant_qos> <discovery> <!--

The initial_peers list are those "addresses" to which the middleware will send discovery announcements.

--> <initial_peers>

<element>4@builtin.shmem://</element> </initial_peers>

<!--

The multicast_receive_addresses list identifies where the DomainParticipant listens for multicast announcements from others. Set this list to an empty value to disable listening over multicast.

--> <multicast_receive_addresses>

<!-- empty --> </multicast_receive_addresses>

</discovery> <transport_builtin>

<!--

The transport_builtin mask identifies which builtin transports the domain participant uses. The default value is UDPv4 | SHMEM, so set this mask to SHMEM to prevent other nodes from initiating communication with this node via UDPv4.

--> <mask>SHMEM</mask>

</transport_builtin>

...

</participant_qos>

For more information, please see the RTI Core Libraries and Utilities Platform Notes, User’s Manual, and the API Reference HTML documentation (from the main page, select Modules, Infrastruc- ture Module, QoS Policies, DISCOVERY).

4.2Customizing Behavior: QoS Configuration

Almost every object in the Connext API is associated with QoS policies that govern its behavior. These policies govern everything from the amount of memory the object may use to store incoming or outgoing data, to the degree of reliability required, to the amount of meta-traffic that the middleware will send on the network, and many others. The following is a short sum- mary of just a few of the available policies:

4-3

Customizing Behavior: QoS Configuration

Reliability and Availability for Consistent Behavior with Critical Data:

❏Reliability: Specifies whether or not the middleware will deliver data reliably. The reliabil- ity of a connection between a DataWriter and DataReader is entirely user configurable. It can be done on a per DataWriter-DataReader connection.

For some use cases, such as the periodic update of sensor values to a GUI displaying the value to a person, best-effort delivery is often good enough. It is the fastest, most effi- cient, and least resource-intensive (CPU and network bandwidth) method of getting the newest/latest value for a topic from DataWriters to DataReaders. But there is no guarantee that the data sent will be received. It may be lost due to a variety of factors, including data loss by the physical transport such as wireless RF or Ethernet.

However, there are data streams (topics) in which you want an absolute guarantee that all data sent by a DataWriter is received reliably by DataReaders. This means that the mid- dleware must check whether or not data was received, and repair any data that was lost by resending a copy of the data as many times as it takes for the DataReader to receive the data.

❏History: Specifies how much data must be stored by the middleware for the DataWriter or DataReader. This QoS policy affects the Reliability and Durability QoS policies.

When a DataWriter sends data or a DataReader receives data, the data sent or received is stored in a cache whose contents are controlled by the History QosPolicy. The History QosPolicy can tell the middleware to store all of the data that was sent or received, or only store the last n values sent or received. If the History QosPolicy is set to keep the last n values only, then after n values have been sent or received, any new data will overwrite the oldest data in the queue. The queue thus acts like a circular buffer of length n.

This QoS policy interacts with the Reliability QosPolicy by controlling whether or not the middleware guarantees that (a) all of the data sent is received (using the KEEP_ALL set- ting of the History QosPolicy) or (b) that only the last n data values sent are received (a reduced level of reliability, using the KEEP_LAST setting of the History QosPolicy). See the Reliability QosPolicy for more information.

Also, the amount of data sent to new DataReaders whose Durability QosPolicy (see below) is set to receive previously published data is controlled by the History QosPolicy.

❏Lifespan: Specifies how long the middleware should consider data sent by a user applica- tion to be valid.

The middleware attaches timestamps to all data sent and received. When you specify a finite Lifespan for your data, the middleware will compare the current time with those timestamps and drop data when your specified Lifespan expires. You can use the Lifes- pan QosPolicy to ensure that applications do not receive or act on data, commands or messages that are too old and have "expired."

❏Durability: Specifies whether or not the middleware will store and deliver previously published data to new DataReaders. This policy helps ensure that DataReaders get all data that was sent by DataWriters, even if it was sent while the DataReader was disconnected from the network. It can increase a system's tolerance to failure conditions.

Fault Tolerance for increased robustness and reduced risk:

❏Liveliness: Specifies and configures the mechanism that allows DataReaders to detect when DataWriters become disconnected or "dead." It can be used during system integra- tion to ensure that systems meet their intended responsiveness specifications. It can also be used during run time to detect possible losses of connectivity.

4-4

Compact, Type-Safe Data: Programming with Data Types

❏Ownership and Ownership Strength: Along with Ownership Strength, Ownership speci- fies if a DataReader can receive data of a given instance from multiple DataWriters at the same time. By default, DataReaders for a given topic can receive data of all instances from any DataWriter for the same topic. But you can also configure a DataReader to receive data of a given instance from only one DataWriter at a time. The DataWriter with the highest Ownership Strength value will be the owner of the instance and the one whose data is delivered to DataReaders. Data of that instance sent by all other DataWriters with lower Ownership Strength will be dropped by the middleware.

When the DataWriter with the highest Ownership strength loses its liveliness (as con- trolled by the Liveliness QosPolicy) or misses a deadline (as controlled by the Deadline QosPolicy) or whose application quits, dies, or otherwise disconnects, the middleware will change ownership of the topic to the DataWriter with the highest Ownership Strength from the remaining DataWriters. This QoS policy can help you build systems that have redundant elements to safeguard against component or application failures. When systems have active and hot standby components, the Ownership QosPolicy can be used to ensure that data from standby applications are only delivered in the case of the failure of the primary.

Built-in Support for Periodic Data:

❏Deadline: For a DataReader, this QoS specifies the maximum expected elapsed time between arriving data samples. For a DataWriter, it specifies a commitment to publish samples with no greater than this elapsed time between them.

This policy can be used during system integration to ensure that applications have been coded to meet design specifications. It can be used during run time to detect when sys- tems are performing outside of design specifications. Receiving applications can take appropriate actions to prevent total system failure when data is not received in time. For topics on which data is not expected to be periodic, the deadline period should be set to an infinite value.

You can specify an object's QoS two ways: (a) programmatically, in your application's source code or (b) in an XML configuration file. The same parameters are available, regardless of which way you choose. For complete information about all of the policies available, see Chapter 4 in the RTI Core Libraries and Utilities User's Manual or see the API Reference HTML documentation.

The examples covered in this document are intended to be configured with XML files. You can find several example configurations, called profiles, in the directory $NDDSHOME/example/ QoS. The easiest way to use one of these profile files is to either set the environment variable NDDS_QOS_PROFILES to the path of the file you want, or copy that file into your current working directory with the file name USER_QOS_PROFILES.xml before running your applica- tion.

4.3Compact, Type-Safe Data: Programming with Data Types

How data is stored or laid out in memory can vary from language to language, compiler to com- piler, operating system to operating system, and processor to processor. This combination of lan- guage/compiler/operating system/processor is called a platform. Any modern middleware must be able to take data from one specific platform (say C/gcc 3.2.2/Solaris/Sparc) and trans- parently deliver it to another (for example, Java/JDK 1.6/Windows XP/Pentium). This process is commonly called serialization/deserialization or marshalling/demarshalling. Messaging products have typically taken one of two approaches to this problem:

4-5

Compact, Type-Safe Data: Programming with Data Types

❏Do nothing. With this approach, the middleware does not provide any help and user code must take into account memory-layout differences when sending messages from one platform to another. The middleware treats messages as an opaque buffer of bytes. The JMS BytesMessage is an example of this approach.

❏Send everything, every time. Self-describing messages are at the opposite extreme, and embed full reflective information, including data types and field names, with each mes- sage. The JMS MapMessage and the messages in TIBCO Rendezvous are examples of this approach.

The “do nothing” approach is lightweight on its surface but forces you, the user of the middle- ware API, to consider all data encoding, alignment, and padding issues. The “send everything” alternative results in large amounts of redundant information being sent with every packet, impacting performance.

Figure 4.1 Self-Describing Messages vs. Type Definitions

Connext exchanges data type definitions, such as field names and types, once at application start- up time. This increases performance and reduces bandwidth consumption compared to the conventional approach, in which each message is self-describing and thus includes a substantial amount of metadata along with the actual data.

Connext takes an intermediate approach. Just as objects in your application program belong to some data type, data samples sent on the same Topic share a data type. This type defines the fields that exist in the data samples and what their constituent types are; users in the aerospace and defense industries will recognize such type definitions as a form of Interface Definition Doc- ument (IDD). Connext stores and propagates this meta-information separately from the individ-

4-6

Compact, Type-Safe Data: Programming with Data Types

ual data samples, allowing it to propagate samples efficiently while handling byte ordering and alignment issues for you.

Table 4.1

Example IDD

0

8

16

24

31

Message ID: unsigned short

Track ID: unsigned short

Position X: short

Position Y: short

Identity:

Position Z: short

Track Type: unsigned short :12

unsigned

short :4

Speed: float

This example IDD shows one legacy approach to type definition. RTI supports multiple standard type definition formats that are machine readable as well as human readable.

With RTI, you have a number of choices when it comes to defining and using data types. You can choose one of these options, or you can mix and match them—these approaches interoper- ate with each other and across programming languages and platforms. So, your options are:

❏Use the built-in types. If a message is simply a string or a buffer of bytes, you can use RTI's built-in types, described in Using Built-in Types (Section 4.3.1).

❏Define a type at compile-time using a language-independent description language and the RTI code generator, rtiddsgen, as described in Using Types Defined at Compile Time (Section 4.3.2). This approach offers the strongest compile-time type safety.

Whereas in-house middleware implementation teams often define data formats in word processing or spreadsheet documents and translate those formats into application code by hand, RTI relies on standard type definition formats that are both human- and machine-readable and generates code in compliance with open international standards. The code generator accepts data-type definitions in a number of formats to make it easy to integrate Connext with your development processes and IT infrastructure:

•OMG IDL. This format is a standard component of both the DDS and CORBA specifi- cations. It describes data types with a C++-like syntax. This format is described in Chapter 3 of the RTI Core Libraries and Utilities User’s Manual.

•XML schema (XSD), whether independent or embedded in a WSDL file. XSD may be the format of choice for those using Connext alongside or connected to a web services infrastructure. This format is described in Chapter 3 of the RTI Core Libraries and Utili- ties User’s Manual.

•XML in a DDS-specific format. This XML format is terser, and therefore easier to read and write by hand, than an XSD file. It offers the general benefits of XML—extensibil- ity and ease of integration—while fully supporting DDS-specific data types and con- cepts. This format is described in Chapter 15 of the RTI Core Libraries and Utilities User’s Manual.

❏Define a dynamic type programmatically at run time.1 This method may be appropriate for applications with dynamic data description needs: applications for which types change frequently or cannot be known ahead of time. It allows you to use an API similar to those of Tibco Rendezvous or JMS MapMessage to manipulate messages without sacri- ficing efficiency. It is described in Running with Dynamic Types (Section 4.3.3).

The following sections of this document describe each of these models.

1. Dynamic types are not supported when using the separate add-on product, RTI Ada 2005 Language Support.

4-7

Compact, Type-Safe Data: Programming with Data Types

4.3.1Using Built-in Types

Connext provides a set of standard types that are built into the middleware. These types can be used immediately. The supported built-in types are String, KeyedString, Octets, and Keyed- Octets. (On Java and .NET platforms, the latter two types are called Bytes and KeyedBytes, respectively; the names are different but the data is compatible across languages.) String and KeyedStrings can be used to send variable-length strings of single-byte characters. Octets and KeyedOctets can be used to send variable-length arrays of bytes.

These built-in types may be sufficient if your data-type needs are simple. If your data is more complex and highly structured, or you want Connext to examine fields within the data for filter- ing or other purposes, this option may not be appropriate, and you will need to take additional steps to use compile-time types (see Using Types Defined at Compile Time (Section 4.3.2)) or dynamic types (see Running with Dynamic Types (Section 4.3.3)).

4.3.2Using Types Defined at Compile Time

In this section, we define a type at compile time using a language-independent description and the RTI code generator, rtiddsgen.

The code generator accepts data-type definitions in a number of formats, such as OMG IDL, XML Schema (XSD), and a DDS-specific format of XML. This makes it easy to integrate Connext with your development processes and IT infrastructure. In this chapter, we will define types using IDL. (In case you would like to experiment with a different format, rtiddsgen can convert from any supported format to any other: simply pass the arguments -convertToIdl, -convert- ToXml, -convertToXsd, or -convertToWsdl.)

As described in the Release Notes, some platforms are supported as both a host and a target, while others are only supported as a target. The rtiddsgen tool must be run on a computer that is supported as a host. For target-only platforms, you will need to run rtiddsgen and build the application on a separate host computer.

The following sections will take your through the process of generating example code from your own data type.

4.3.2.1Generating Code with rtiddsgen

Don't worry about how types are defined in detail for now (we cover it in Chapter 3 of the RTI Core Libraries and Utilities User's Manual). For this example, just copy and paste the following into a new file, HelloWorld.idl.

const long HELLO_MAX_STRING_SIZE = 256;

struct HelloWorld { string<HELLO_MAX_STRING_SIZE> message;

};

Next, we will invoke the rtiddsgen code-generator, which can be found in the $NDDSHOME/ scripts directory that should already be on your path, to create definitions of your data type in a target programming language, including logic to serialize and deserialize instances of that type for transmission over the network. Then we will build and run the generated code.

For a complete list of the arguments rtiddsgen understands, and a brief description of each of them, run it with the -help argument. More information about rtiddsgen, including its command- line parameters and the list of files it creates, can be found in Section 3.7 of the RTI Core Libraries and Utilities User's Manual.

4-8

Compact, Type-Safe Data: Programming with Data Types

4.3.2.1.2Instructions for C++

Generate C++ code from your IDL file with the following command (replace the architecture name i86Linux2.6gcc3.4.3 with the name of your own architecture):

> rtiddsgen -ppDisable	\
-language C++	\
-example i86Linux2.6gcc3.4.3 \
-replace	\
HelloWorld.idl

The generated code publishes identical data samples and subscribes to them, printing the received data to the terminal. Edit the code to modify each data sample before it's published: Open HelloWorld_publisher.cxx. In the code for publisher_main(), locate the "for" loop and add the bold line seen below, which puts "Hello World!" and a consecutive number in each sam- ple that is sent.1

for (count=0; (sample_count == 0) || (count < sample_count); ++count) {

printf("Writing HelloWorld, count %d\n", count);

/* Modify the data to be sent here */ sprintf(instance->message, "Hello World! (%d)", count);

retcode = HelloWorld_writer->write(*instance, instance_handle); if (retcode != DDS_RETCODE_OK) {

printf("write error %d\n", retcode);

}

NDDSUtility::sleep(send_period);

}

4.3.2.1.3Instructions for Java

Generate Java code from your IDL file with the following command2 (replace the architecture name i86Linux2.6gcc3.4.3jdk with the name of your own architecture):

> rtiddsgen -ppDisable	\
-language Java	\
-example i86Linux2.6gcc3.4.3jdk \
-replace	\
HelloWorld.idl

The generated code publishes identical data samples and subscribes to them, printing the received data to the terminal. Edit the code to modify each data sample before it's published: Open HelloWorldPublisher.java. In the code for publisherMain(), locate the "for" loop and add

1.If you are using Visual Studio 2005 or Visual Studio 2008, consider using sprintf_s instead of sprintf: sprintf_s(instance->msg, 128, "Hello World! (%d)", count);

2.The argument -ppDisable tells the code generator not to attempt to invoke the C preprocessor (usually cpp on UNIX systems and cl on Windows systems) on the IDL file prior to generating code. In this case, the IDL file con- tains no preprocessor directives, so no preprocessing is necessary. However, if the preprocessor executable is already on your system path (on Windows systems, running the Visual Studio script vcvars32.bat will do this for you) you can omit this argument.

4-9

Compact, Type-Safe Data: Programming with Data Types

the bold line seen below, which puts "Hello World!" and a consecutive number in each sample that is sent.

for (int count = 0; sampleCount == 0) || (count < sampleCount); ++count) {

System.out.println("Writing HelloWorld, count " + count);

/* Modify the instance to be written here */ instance.msg = "Hello World! (" + count + ")";

/* Write data */

writer.write(instance, InstanceHandle_t.HANDLE_NIL); try {

Thread.sleep(sendPeriodMillis); } catch (InterruptedException ix) {

System.err.println("INTERRUPTED");

break;

}

}

4.3.2.1.1Instructions for Ada

Generate Ada code from your IDL file with the following command (replace the architecture name i86Linux2.6gcc4.1.2 with the name of your own architecture):

rtiddsgen -ppDisable -language Ada -example i86Linux2.6gcc4.1.2 \ -replace HelloWorld.idl

Notes:

❏Ada support requires a separate add-on product, Ada 2005 Language Support.

❏The generated publisher and subscriber source files are under the samples directory. There are two generated project files: one at the top level and one in the samples direc- tory. The project file in the samples directory, samples/helloworld-samples.gpr, should be the one that you will use to compile the example.

❏The generated Ada project files need two directories to compile a project: .obj and bin. If your Ada IDE does not automatically create these directories, you will need to create them outside the Ada IDE, in both the top-level directory and in samples directory.

for the “Count in 0 .. Sample_Count “loop

Put_Line ("Writing HelloWorld, count " & Count'Img); declare

Msg : DDS.String := DDS.To_DDS_String ("Hello World! (" & Count'Img & ")");

begin

if Instance.message /= DDS.NULL_STRING then Finalize (Instance.message);

end if;

Instance.message.data := Msg.data;

end;

HelloWorld_Writer.Write (Instance_Data => Instance, Handle => Instance_Handle'Unchecked_Access);

delay Send_Period; end loop;

if Instance.message /= DDS.NULL_STRING then Finalize (Instance.message);

end if;

4-10

Compact, Type-Safe Data: Programming with Data Types

4.3.2.2Building the Generated Code

You have now defined your data type, generated code for it, and customized that code. It's time to compile the example applications.

4.3.2.2.1Building a Generated C, C++, or .NET Example on Windows Systems

With the NDDSHOME environment variable set, start Visual Studio and open the rtiddsgen-

generated workspace (.dsw) or solution object (.sln) file. Select the Win32 Release configuration in the Build toolbar in Visual Studio, or the Standard toolbar in Visual Studio .NET1. From the

Build menu, select Build Solution. This will build two projects: <IDL name>_publisher and

<IDL name>_subscriber2.

4.3.2.2.2Building a Generated Ada Example on a Linux System

Ada support requires a separate add-on product, Ada 2005 Language Support. Use the generated Ada project file to compile an example on any system.

Note: The generated project file assumes the correct version of the compiler is already on your path, NDDSHOME is set, and $NDDSHOME/lib/gnat is in your ADA_PROJECT_PATH.

gmake -f makefile_HelloWorld_<architecture>

After compiling the Ada example, you will find the application executables in the directory, samples/bin. The build command in the generated makefile uses the static release versions of the Connext libraries. To select dynamic or debug versions of the libraries, change the Ada compiler variables LIBRARY_TYPE and NDDS_BUILD in the build command in the makefile to build with the desired version of the libraries. For example, if the application must be compiled with the relocatable debug version of the libraries, compile with the following command:

gprbuild -p -P samples/helloworld-samples.gpr -XOS=Linux \ -XLIBRARY_TYPE=relocatable -XNDDS_BUILD=debug -XARCH=${ARCH}

4.3.2.2.3Building a Generated Example on Other Platforms

Use the generated makefile to compile a C or C++ example on a UNIX-based system or a Java example on any system. Note: the generated makefile assumes the correct version of the com- piler is already on your path and that NDDSHOME is set.

gmake -f makefile_HelloWorld_<architecture>

After compiling the C or C++ example, you will find the application executables in a directory objs/<architecture>.

The generated makefile includes the static release versions of the Connext libraries. To select dynamic or debug versions of the libraries, edit the makefile to change the library suffixes. Gener- ally, Connext uses the following convention for library suffixes: "z" for static release, "zd" for static debug, none for dynamic release, and "d" for dynamic debug. For a complete list of the required libraries for each configuration, see the RTI Core Libraries and Utilities Platform Notes.

1.RTI Connext .NET language binding is currently supported for C# and C++/CLI.

2.There is one exception. For C# users: Select the Mixed Platforms Release configuration in the Standard toolbar in Visual Studio (.NET) 2005. Select Build Solution from the Build menu. This will build 3 projects:

<IDL name>_type, <IDL name>_publisher and <IDL name>_subscriber. Note for Visual Studio 2008 users: Although the project files generated by rtiddsgen are for Visual Studio 2005, you can open them with Visual Studio 2008. Visual 2008 will launch an upgrade wizard to convert the files to Visual Studio 2008 project files. Once the upgrade is done, the rest of the steps are the same.

4-11

Compact, Type-Safe Data: Programming with Data Types

For example, to change a C++ makefile from using static release to dynamic release libraries, change this directive:

LIBS = -L$(NDDSHOME)/lib/<architecture> \

-lnddscppz -lnddscz -lnddscorez $(syslibs_<architecture>)

to:

LIBS = -L$(NDDSHOME)/lib/<architecture> \

-lnddscpp -lnddsc -lnddscore $(syslibs_<architecture>)

4.3.2.3Running the Example Applications

Run the example publishing and subscribing applications and see them communicate:

4.3.2.3.1Running the Generated C++ Example

First, start the subscriber application, HelloWorld_subscriber:

./objs/<architecture>/HelloWorld_subscriber

In this command window, you should see that the subscriber wakes up every four seconds to

print a message:

HelloWorld subscriber sleeping for 4 sec...

HelloWorld subscriber sleeping for 4 sec...

HelloWorld subscriber sleeping for 4 sec...

Next, open another command prompt window and start the publisher application,

HelloWorld_publisher. For example:

./objs/<architecture>/HelloWorld_publisher

In this second (publishing) command window, you should see:

Writing HelloWorld, count 0

Writing HelloWorld, count 1

Writing HelloWorld, count 2

Look back in the first (subscribing) command window. You should see that the subscriber is now receiving messages from the publisher:

HelloWorld subscriber sleeping for 4 sec...

msg: “Hello World! {0}“

HelloWorld subscriber sleeping for 4 sec...

msg: “Hello World! {1}“

HelloWorld subscriber sleeping for 4 sec...

msg: “Hello World! {2}“

4.3.2.3.2Running the Generated Java Example

You can run the generated applications using the generated makefile. On most platforms, the

generated makefile assumes the correct version of java is already on your path and that the NDDSHOME environment variable is set1.

First, run the subscriber:

gmake -f makefile_HelloWorld_<architecture> HelloWorldSubscriber

1.One exception in LynxOS; see the chapter Getting Started on Embedded UNIX-like Systems in the Getting Started Guide, Addendum for Embedded Platforms.

4-12

Compact, Type-Safe Data: Programming with Data Types

In this command window, you should see that the subscriber wakes up every four seconds to print a message:

HelloWorld subscriber sleeping for 4 sec...

HelloWorld subscriber sleeping for 4 sec...

HelloWorld subscriber sleeping for 4 sec...

Next, run the publisher:

gmake -f makefile_HelloWorld_<architecture> HelloWorldPublisher

In this second (publishing) command window, you should see:

Writing HelloWorld, count 0

Writing HelloWorld, count 1

Writing HelloWorld, count 2

Look back in the first (subscribing) command window. You should see that the subscriber is now receiving messages from the publisher:

HelloWorld subscriber sleeping for 4 sec...

msg: “Hello World! {0}“

HelloWorld subscriber sleeping for 4 sec...

msg: “Hello World! {1}“

HelloWorld subscriber sleeping for 4 sec...

msg: “Hello World! {2}“

4.3.2.3.3Running the Generated Ada Example

Ada support requires a separate add-on product, Ada 2005 Language Support.

First, start the subscriber application, helloworld_idl_file-helloworld_subscriber::

./samples/bin/helloworld_idl_file-helloworld_subscriber

In this command window, you should see that the subscriber wakes up every four seconds to

print a message:

HelloWorld subscriber sleeping for 4.000000000 sec.

HelloWorld subscriber sleeping for 4.000000000 sec.

HelloWorld subscriber sleeping for 4.000000000 sec.

Next, open another command prompt window and start the publisher application, helloworld_idl_file-helloworld_publisher. For example:

./samples/bin/helloworld_idl_file-helloworld_publisher

In this second (publishing) command window, you should see:

Writing HelloWorld, count 0

Writing HelloWorld, count 1

Writing HelloWorld, count 2

Look back in the first (subscribing) command window. You should see that the subscriber is now receiving messages from the publisher:

HelloWorld subscriber sleeping for 4 sec...

msg: “Hello World! {0}“

HelloWorld subscriber sleeping for 4 sec...

msg: “Hello World! {1}“

HelloWorld subscriber sleeping for 4 sec...

msg: “Hello World! {2}“

4-13

Compact, Type-Safe Data: Programming with Data Types

4.3.3Running with Dynamic Types

Note: Dynamic types are not supported when using Ada 2005 Language Support

This method may be appropriate for applications for which the structure (type) of messages changes frequently or for deployed systems in which newer versions of applications need to interoperate with existing applications that cannot be recompiled to incorporate message-type changes.

As your system evolves, you may find that your data types need to change. And unless your system is relatively small, you may not be able to bring it all down at once in order to modify them. Instead, you may need to upgrade your types one component at a time-or even on the fly, without bringing any part of the system down.

While covering dynamic types is outside the scope of this chapter, you can learn more about the subject in Chapter 3 of the RTI Core Libraries and Utilities User's Manual. You can also view and run the Hello World example code located in $NDDSHOME/example/<language>/ Hello_dynamic/src.

4-14

Chapter 5 Design Patterns for Rapid Development

In this chapter, you will learn how to implement some common functional design patterns. As you have learned, one of the advantages to using Connext is that you can achieve significantly different functionality without changing your application code simply by updating the XML- based Quality of Service (QoS) parameters.

In this chapter, we will look at a simple newspaper example to illustrate these design patterns. Newspaper distribution has long been a canonical example of publish-subscribe communica- tion, because it provides a simple metaphor for real-world problems in a variety of industries.

A radar tracking system and a market data distribution system share many features with the news subscriptions we are all familiar with: many-to-many publish-subscribe communication with certain quality-of-service requirements.

In a newspaper scenario (provided in an example called News example for all languages, a news publishing application distributes articles from a variety of news outlets—CNN, Bloom- berg, etc.—on a periodic basis. However, the period differs from outlet to outlet. One or more news subscribers poll for available articles, also on a periodic basis, and print out their contents. Once published, articles remain available for a period of time, during which subscribing appli- cations can repeatedly access them if they wish. After that time has elapsed, the middleware will automatically expire them from its internal cache.

This chapter describes Building and Running the Code Examples (Section 5.1) and includes the following design patterns:

❏Subscribing Only to Relevant Data (Section 5.2)

❏Accessing Historical Data when Joining the Network (Section 5.3)

5-1

Building and Running the Code Examples

❏Caching Data Within the Middleware (Section 5.4)

❏Receiving Notifications When Data Delivery Is Late (Section 5.5)

You can find more examples at http://www.rti.com/examples. This page contains example code snippets on how to use individual features, examples illustrating specific use cases, as well as performance test examples.

Next Chapter —> Chapter 6: Design Patterns for High Performance

5.1Building and Running the Code Examples

Source code for the News example is located in the directory ${NDDSHOME}/example/<lan- guage>/News.

The example performs these steps:

1.Parses the command-line arguments.

2.Loads the quality-of-service (QoS) file, USER_QOS_PROFILES.xml, from the current working directory. The middleware does this automatically; you will not see code in the example to do this. For more information on how to use QoS profiles, see the chapter on "Configuring QoS with XML" in the RTI Core Libraries and Utilities User's Manual.

3.On the publishing side, sends news articles periodically.

4.On the subscribing side, receives these articles and print them out periodically.

The steps for compiling and running the program are similar to those described in Building and Running “Hello, World” (Section 3.1). As with the Hello World example, Java users should use the build and run command scripts in the directory ${NDDSHOME}/example/JAVA/News. Non-Java Windows developers will find the necessary Microsoft Visual Studio solution files in the subdirectory ${NDDSHOME}/example/<language>/News/win32. Appropriate makefiles for building the C, C++, or Ada1 examples on UNIX platforms are located in the subdirectory ${NDDSHOME}/example/<language>/News/make.

The News example combines the publisher and subscriber in a single program, which is run from the ${NDDSHOME}/example/<language>/News directory with the argument pub or sub to select the desired behavior. When running with default arguments, you should see output similar to that shown in Figure 5.1. To see additional command-line options, run the program without any arguments.

Both the publishing application and the subscribing application operate in a periodic fashion:

❏The publishing application writes articles from different news outlets at different rates. Each of these articles is numbered consecutively with respect to its news outlet. Every two seconds, the publisher prints a summary of what it wrote in the previous period.

❏The subscribing application polls the cache of its DataReader every two seconds and prints the data it finds there. (Many real-world applications will choose to process data as soon as it arrives rather than polling for it. For more information about the different ways to read data, select Modules, Programming How-To’s, DataReader Use Cases in

1. Ada support requires a separate add-on product, Ada 2005 Language Support.

5-2

Subscribing Only to Relevant Data

the API Reference HTML documentation. In this case, periodic polling makes the behav- ior easy to illustrate.) Along with each article it prints, it includes the time at which that article was published and whether that article has been read before or whether it was cached from a previous read.

By default, both the publishing and subscribing applications run for 20 seconds and then quit. (To run them for a different number of seconds, use the -r command-line argument.) Figure 5.1 shows example output using the default run time and a domain ID of 13 (set with the command- line option, -d 13).

Figure 5.1 Example Output for Both Applications

Publishing Application

Subscribing Application

5.2Subscribing Only to Relevant Data

From reading Chapter 3, you already understand how to subscribe only to the topics in which you’re interested. However, depending on your application, you may be interested in only a fraction of the data available on those topics. Extraneous, uninteresting, or obsolete data puts a drain on your network and CPU resources and complicates your application logic.

Fortunately, Connext can perform much of your filtering and data reduction for you. Data reduc- tion is a general term for discarding unnecessary or irrelevant data so that you can spend you time processing the data you care about. You can define the set of data that is relevant to you based on:

❏Its content. Content-based filters can examine any field in your data based on a variety of criteria, such as whether numeric values meet various equality and inequality relation- ship or whether string values match certain regular expression patterns. For example, you may choose to distribute a stream of stock prices using a single topic, but indicate that you’re interested only in the price of IBM, and only when that price goes above $20.

5-3

Subscribing Only to Relevant Data

❏How old it is. You can indicate that data is relevant for only a certain period of time (its lifespan) and/or that you wish to retain only a certain number of data samples (a history depth). For example, if you are interested in the last ten articles from each news outlet, you can set the history depth to 10.

❏How fast it’s updated. Sometimes data streams represent the state of a moving or chang- ing object—for example, an application may publish the position of a moving vehicle or the changing price of a certain financial instrument. Subscribing applications may only be able to—or interested in—processing this data at a certain maximum rate. For exam- ple, if the changing state is to be plotted in a user interface, a human viewer is unlikely to be able to process more than a couple of changes per second. If the application attempts to process updates any faster, it will only confuse the viewer and consume resources unnecessarily.

5.2.1Content-Based Filtering

A DataReader can filter incoming data by subscribing not to a given Topic itself, but to a Content- FilteredTopic that associates the Topic with an SQL-like expression that indicates which data sam- ples are of interest to the subscribing application. Each subscribing application can specify its own content-based filter, as it desires; publishing applications’ code does not need to change to allow subscribers to filter on data contents.

5.2.1.1Implementation

The C++ code looks like this:

DDSContentFilteredTopic cft = participant->create_contentfilteredtopic( cftName.c_str(),

topic, contentFilterExpression.c_str(), noFilterParams);

if (cft == NULL) {

throw std::runtime_error("Unable to create ContentFilteredTopic");

}

The Topic and ContentFilteredTopic classes share a base class: TopicDescription.

The variable contentFilterExpression in the above example code is a SQL expression; see imme- diately below.

The corresponding code in Java looks like this:

ContentFilteredTopic cft = participant.create_contentfilteredtopic( topic.get_name() + " (filtered)",

topic,

contentFilterExpression,

null); if (cft == null) {

throw new IllegalStateException(

"Unable to create ContentFilteredTopic");

}

5-4

Subscribing Only to Relevant Data

The Ada1 code to create a ContentFilteredTopic looks like this:

declare

cft : DDS.ContentFilteredTopic.Ref_Access;

cftName : DDS.String := DDS.To_DDS_String (DDS.To_Standard_String (topic.Get_Name) & " (filtered)");

begin

cft := participant.Create_Contentfilteredtopic

(cftName,

topic,

contentFilterExpression,

null);

DDS.Finalize (cftName);

if cft = null then

Put_Line (Standard_Error,

"Unable to create ContentFilteredTopic");

return; end if;

end;

In Ada, ContentFilteredTopic class inherits from TopicDescription while Topic has a function ‘As_TopicDescription’ that is useful to create a DataReader where a TopicDescription is needed.

5.2.1.2Running & Verifying

The following shows a simple filter with which the subscribing application indicates it is inter- ested only in news articles from CNN; you can specify such a filter with the –f or --filterExpres- sion command-line argument, like this in the Java example:

>./run.sh sub -f "key='CNN' OR key='Reuters'"

This example uses the built-in KeyedString data type. This type has two fields: key, a string field that is the type’s only key field, and value, a second string. This example uses the key field to store the news outlet name and the value field to store the article text. The word “key” in the content filter expression “key='CNN'” refers to the field’s name, not the fact that it is a key field; you can also filter on the value field if you like. In your own data types, you will use the name(s) of the fields you define. For example output, see Figure 5.2 on page 5-6.

You can find more detailed information in the API Reference HTML documentation under

Modules, RTI Connext API Reference, Queries and Filters Syntax. For Ada1, open doc/html/ api_ada/index.html and look under RTI Connext API Reference, DDSQueryAndFilterSyntax- Module.

5.2.2Lifespan and History Depth

One of the most common ways to reduce data is to indicate for how long a data sample remains valid once it has been sent. You can indicate such a contract in one (or both) of two ways: in terms of a number of samples to retain (the history depth) and the elapsed time period during which a sample should be retained (the lifespan). For example, you may be interested in only the most recent data value (the so-called “last values”) or in all data that has been sent during the last sec- ond.

The history depth and lifespan, which can be specified declaratively with QoS policies (see below), apply both to durable historical data sent to late-joining subscribing applications as well as to current subscribing applications. For example, suppose a subscribing application has set history depth = 2, indicating that it is only interested in the last two samples. A publishing appli-

1.Ada support requires a separate add-on product, Ada 2005 Language Support.

5-5

Subscribing Only to Relevant Data

Figure 5.2 Using a Content-based Filter

Publishing Application

Subscribing Application

cation sends four data samples in close succession, and the middleware on the subscribing side receives them before the application chooses to read them from the middleware. When the sub- scribing application does eventually call DataReader::read(), it will see only samples 3 and 4; samples 1 and 2 will already have been overwritten. If an application specifies both a finite his- tory depth and a finite lifespan, whichever limit is reached first will take effect.

5.2.2.1Implementation

History depth is part of the History QoS policy. The lifespan is specified with the Lifespan QoS policy. The History QoS policy applies independently to both the DataReader and the DataWriter; the values specified on both sides of the communication need not agree. The Lifespan QoS pol- icy is specified only on the DataWriter, but it is enforced on both sides: the DataReader will enforce the lifespan indicated by each DataWriter it discovers for each sample it receives from that DataWriter.

For more information about these QoS policies, consult the API Reference HTML documenta-

tion. Open ReadMe.html and select a programming language, then select Modules, RTI Con- next API Reference, Infrastructure, QoS Policies. For Ada1: open doc/html/api_ada/ index.html and select Infrastructure Module, DDSQosTypesModule.

You can specify these QoS policies either in your application code or in one or more XML files. Both mechanisms are functionally equivalent; the News example provided for C, C++, Java, and Ada uses XML files. For more information about this mechanism, see the chapter on “Configur- ing QoS with XML” in the RTI Core Libraries and Utilities User’s Manual.

1. Ada support requires a separate add-on product, Ada 2005 Language Support.

5-6

Subscribing Only to Relevant Data

5.2.2.1.1History Depth

The DDS specification, which Connext implements, recognizes two “kinds” of history: KEEP_ALL and KEEP_LAST. KEEP_ALL history indicates that the application wants to see every data sample, regardless of how old it is (subject, of course, to lifespan and other QoS poli- cies). KEEP_LAST history indicates that only a certain number of back samples are relevant; this number is indicated by a second parameter, the history depth. (The depth value, if any, is ignored if the kind is set to KEEP_ALL.)

To specify a default History policy of KEEP_LAST and a depth of 10 in a QoS profile:

<history> <kind>KEEP_LAST_HISTORY_QOS</kind> <depth>10</depth>

</history>

You can see this in the News example (provided for C, C++, Java, and Ada1) in the file

USER_QOS_PROFILES.xml.

Lifespan Duration

The Lifespan QoS policy contains a field duration that indicates for how long each sample remains valid. The duration is measured relative to the sample’s reception time stamp, which means that it doesn’t include the latency of the underlying transport.

To specify a Lifespan duration of six seconds:

<lifespan>

<duration>

<sec>6</sec>

<nanosec>0</nanosec>

</duration>

</lifespan>

You can see this in the News example (provided for C, C++, Java, and Ada1) in the file

USER_QOS_PROFILES.xml.

5.2.2.2Running & Verifying

The News example never takes samples from the middleware, it only reads samples, so data sam- ples that have already been viewed by the subscribing application remain in the middleware’s internal cache until they are expired by their history depth or lifespan duration contract. These previously viewed samples are displayed by the subscribing application as “cached” to make them easy to spot. See how “CNN” articles are expired:

It's important to understand that the data type used by this example is keyed on the name of the news outlet and that the history depth is enforced on a per-instance basis. You can see the effect in Figure 5.3: even though Reuters publishes articles faster than CNN, the Reuters articles do not "starve out" the CNN articles; each outlet gets its own depth samples. (Figure 5.3 only shows articles from CNN and Reuters, because that makes it easier to fit

more data on the page. If you run the example without a content filter, you will see the same effect across all news outlets.).

Next, we will change the history depth and lifespan duration in the file USER_QOS_PROFILES.xml to see how the example's output changes.

Set the history depth to one.

1. Ada support requires a separate add-on product, Ada 2005 Language Support.

5-7

Subscribing Only to Relevant Data

Figure 5.3 Using History and Lifespan

Article 1 initially received from Reuters and CNN

CNN articles still available

Reuters articles 1-5 expired

due to history depth (10 samples)

CNN articles 1 & 2 expired due to lifespan (6 seconds)

Run the example again with the content-based filter, shown here in the Java example:

> ./run.sh sub -f "key='CNN' OR key='Reuters'"

You will only see articles from CNN and Reuters, and only the last value published for each (as of the end of the two-second polling period).

Now set the history depth to 10 and decrease the lifespan to three seconds.

With these settings and at the rates used by this example, the lifespan will always take effect before the history depth. Run the example again, this time without a content filter. Notice how the subscribing application sees all of the data that was published during the last two-second period as well as the data that was published in the latter half of the previous period.

5-8

Subscribing Only to Relevant Data

Reduce the lifespan again, this time to one second.

If you run the example now, you will see no cached articles at all. Do you understand why? The subscribing application’s polling period is two seconds, so by the time the next poll comes, everything seen the previous time has expired.

Try changing the history depth and lifespan duration in the file USER_QOS_PROFILES.xml to see how the example’s output changes.

5.2.3Time-Based Filtering

A time-based filter allows you to specify a minimum separation between the data samples your subscribing application receives. If data is published faster than this rate, the middleware will discard the intervening samples.

Such a filter is most often used to down-sample high-rate periodic data (see also Receiving Noti- fications When Data Delivery Is Late (Section 5.5)) but it can also be used to limit data rates for aperiodic-but-bursty data streams. Time-based filters have several applications:

❏You can limit data update rates for applications in which rapid updates would be unnec- essary or inappropriate. For example, a graphical user interface should typically not update itself more than a few times each second; more frequent updates can cause flick- ering or make it difficult for a human operator to perceive the correct values.

❏You can reduce the CPU requirements for less-capable subscribing machines to improve their performance. If your data stream is reliable, helping slow readers keep up can actu- ally improve the effective throughput for all readers by preventing them from throttling the writer.

❏In some cases, you can reduce your network bandwidth utilization, because the writer can transparently discard unwanted data before it is even sent on the network.

5.2.3.1Implementation

Time-based filters are specified using the TimeBasedFilter QoS policy. It only applies to

DataReaders.

For more information about this QoS policy, consult the API Reference HTML documentation.

Open ReadMe.html and select a programming language, then select Modules, RTI Connext API Reference, Infrastructure, QoS Policies. For Ada1: open doc/html/api_ada/index.html and select Infrastructure Module, DDSQosTypesModule.

You can specify the QoS policies either in your application code or in one or more XML files. Both mechanisms are functionally equivalent; the News example uses the XML mechanism. For more information, see the chapter on “Configuring QoS with XML” in the RTI Core Libraries and Utilities User’s Manual.

1. Ada support requires a separate add-on product, Ada 2005 Language Support.

5-9

Subscribing Only to Relevant Data

To specify a time-based filter in a QoS policy:

<!-- <time_based_filter>

<minimum_separation> <sec>1</sec> <nanosec>1</nanosec>

</minimum_separation> </time_based_filter> <deadline>

<period>

<sec>3</sec>

<nanosec>0</nanosec>

</period>

</deadline> -->

You can see this in the file USER_QOS_PROFILES.xml provided with the C, C++, Java, and Ada1 News example (uncomment it to specify a time-based filter).

At the same time we implement a time-based filter in the News example, we increase the dead- line period. See the accompanying comment in the XML file as well as the API Reference HTML documentation for more information about using the Deadline and TimeBasedFilter QoS poli- cies together.

5.2.3.2Running & Verifying

Figure 5.4 shows some of the output after activating the filter:

Figure 5.4 Using a Time-based Filter

Article 1 sent at :31

Article 1 sent at :31

Article 6 sent at :32.

Articles 2-5 filtered.

Article 3 sent at :32.

Article 2 filtered.

Because you set the time-based filter to one second, you will not see more than two updates for any single news outlet in any given period, because the period is two seconds long.

1. Ada support requires a separate add-on product, Ada 2005 Language Support.

5-10

Accessing Historical Data when Joining the Network

5.3Accessing Historical Data when Joining the Network

In Section 5.2, you learned how to specify which data on the network is of interest to your appli- cation. The same QoS parameters can also apply to late-joining subscribers, applications that subscribe to a topic after some amount of data has already been published on that topic. This concept—storing sent data within the middleware—is referred to as durability. You only need to indicate the degree of durability in which you’re interested:

❏Volatile. Data is relevant to current subscribers only. Once it has been acknowledged (if reliable communication has been turned on), it can be removed from the service. This level of durability is the default; if you specify nothing, you will get this behavior.

❏Transient local. Data that has been sent may be relevant to late-joining subscribers (sub- ject to any history depth, lifespan, and content- and/or time-based filters defined). His- torical data will be cached with the DataWriter that originally produced it. Once that writer has been shut down for any reason, intentionally or unintentionally, however, the data will no longer be available. This lightweight level of durability is appropriate for non-critical data streams without stringent data availability requirements.

❏Transient. Data that has been sent may be relevant to late-joining subscribers and will be stored externally to the DataWriter that produced that data. This level of durability requires one or more instances of the RTI Persistence Service on your network. As long as one or more of these persistence service instances is functional, the durable data will con- tinue to be available, even if the original DataWriter shuts down or fails. However, if all instances of the persistence service shut down or fail, the durable data they were main- taining will be lost. This level of durability provides a higher level of fault tolerance and availability than does transient-local durability without the performance or management overhead of a database.

❏Persistent. Data that has been sent may be relevant to late-joining subscribers and will be stored externally to the DataWriter that produced that data in a relational database. This level of durability requires one or more instances of the RTI Persistence Service on your network. It provides the greatest degree of assurance for your most critical data streams, because even if all data writers and all persistence server instances fail, your data can nevertheless be restored and made available to subscribers when you restart the persis- tence service.

As you can see, the level of durability indicates not only whether historical data will be made available to late-joining subscribers; it also indicates the level of fault tolerance with which that data will be made available. Learn more about durability, including the RTI Persistence Service, by reading the chapter on “Mechanisms for Achieving Information Durability and Persistence” in the RTI Core Libraries and Utilities User’s Manual.

5.3.1Implementation

To configure data durability, use the Durability QoS policy on your DataReader and/or DataW- riter. The degrees of durability described in Section 5.3 are represented as an enumerated dura- bility kind.

For more information about this QoS policy, consult the API Reference HTML documentation.

Open ReadMe.html and select the API documentation for your language then select Modules, RTI Connext API Reference, Infrastructure, QoS Policies. For Ada1: open doc/html/api_ada/ index.html and select Infrastructure Module, DDSQosTypesModule.

1.Ada support requires a separate add-on product, Ada 2005 Language Support.

5-11

Caching Data Within the Middleware

You can specify the QoS policies either in your application code or in one or more XML files. Both mechanisms are functionally equivalent; the News example uses the XML mechanism. For more information, see the chapter on “Configuring QoS with XML” in the RTI Core Libraries and Utilities User’s Manual.

Here is an example of how to configure the Durability QoS policy in a QoS profile:

<durability> <kind>TRANSIENT_LOCAL_DURABILITY_QOS</kind>

</durability>

You can see this in the News example (provided for C, C++, Java, and Ada1) in the file

USER_QOS_PROFILES.xml.

The above configuration indicates that the DataWriter should maintain data it has published on behalf of later-joining DataReaders, and that DataReaders should expect to receive historical data when they join the network. However, if a DataWriter starts up, publishes some data, and then shuts down, a DataReader that subsequently starts up will not receive that data.

Durability, like some other QoS policies, has request-offer semantics: the DataWriter must offer a level of service that is greater than or equal to the level of service requested by the DataReader. For example, a DataReader may request only volatile durability, while the DataWriter may offer tran- sient durability. In such a case, the two will be able to communicate. However, if the situation were reversed, they would not be able to communicate.

5.3.2Running & Verifying

Run the NewsPublisher and wait several seconds. Then start the NewsSubscriber. Look at the time stamps printed next to the received data: you will see that the subscribing application receives data that was published before it started up.

Now do the same thing again, but first modify the configuration file by commenting the dura- bility configuration. You will see that the subscribing application does not receive any data that was sent prior to when it joined the network.

5.4Caching Data Within the Middleware

When you receive data from the middleware in your subscribing application, you may be able to process all of the data immediately and then discard it. Frequently, however, you will need to store it somewhere in order to process it later. Since you’ve already expressed to the middleware how long your data remains relevant (see Subscribing Only to Relevant Data (Section 5.2)), wouldn’t it be nice if you could take advantage of the middleware’s own data cache rather than implementing your own? You can.

When a DataReader reads data from the network, it places the samples, in order, into its internal cache. When you’re ready to view that data (either because you received a notification that it was available or because you decided to poll), you use one of two families of methods:

❏take: Read the data from the cache and simultaneously remove it from that cache. Future access to that DataReader’s cache will not see any data that was previously taken from the

cache. This behavior is similar to the behavior of “receive” methods provided by tradi- tional messaging middleware implementations. The generated Ada2 example uses this method.

1.Ada support requires a separate add-on product, Ada 2005 Language Support.

2.Ada support requires a separate add-on product, Ada 2005 Language Support.

5-12

Caching Data Within the Middleware

❏read: Read the data from the cache but leave it in the cache so that it can potentially be read again (subject to any lifespan or history depth you may have specified). The News exam- ple for C, C++ and Java uses this method.

When you read or take data from a DataReader, you can indicate that you wish to access all of the data in the cache, all up to a certain maximum number of samples, all of the new samples that you have never read before, and/or various other qualifiers. If your topic is keyed, you can choose to access the samples of all instances at once, or you can read/take one instance at a time. For more information about keys and instances, see Section 2.2.2, “Samples, Instances, and Keys” in the RTI Core Libraries and Utilities User’s Manual.

5.4.1Implementation

The call to read looks like this in C++:

DDS_ReturnCode_t result = _reader->read(

articles,	//	fill in	data here
articleInfos,	//	fill in	parallel meta-data here
DDS_LENGTH_UNLIMITED, //		any # articles

DDS_ANY_SAMPLE_STATE, DDS_ANY_VIEW_STATE, DDS_ANY_INSTANCE_STATE);

if (result == DDS_RETCODE_NO_DATA) {

// nothing to read; go back to sleep

}

if (result != DDS_RETCODE_OK) {

// an error occurred: stop reading

throw std::runtime_error("A read error occurred: " + result);

}

// Process data...

_reader->return_loan(articles, articleInfos);

It looks like this in Java:

try {
_reader.read(
articles,	// fill	in data here
articleInfos,	// fill	in parallel meta-data here

ResourceLimitsQosPolicy.LENGTH_UNLIMITED, // any # articles

SampleStateKind.ANY_SAMPLE_STATE,

ViewStateKind.ANY_VIEW_STATE,

InstanceStateKind.ANY_INSTANCE_STATE);

//Process data...

}catch (RETCODE_NO_DATA noData) {

//nothing to read; go back to sleep

}catch (RETCODE_ERROR ex) {

//an error occurred: stop reading throw ex;

}finally {

_reader.return_loan(articles, articleInfos);

}

5-13

		Caching Data Within the Middleware
And in Ada1:
	declare
	received_data	: aliased DDS.KeyedString_Seq.Sequence;
	sample_info	: aliased DDS.SampleInfo_Seq.Sequence;
	begin
	reader.Read
	(articles'Access, -- fill in data here
	articleInfos'Access, -- fill in parallel meta-data here
	DDS.LENGTH_UNLIMITED, -- any # articles
		(received_data'Access,
		sample_info'Access,
		DDS.LENGTH_UNLIMITED,
		DDS.ANY_SAMPLE_STATE,
		DDS.ANY_VIEW_STATE,
		DDS.ANY_INSTANCE_STATE);
	-- Process data...
	reader.Return_Loan (articles, articleInfos);
	for i in 1 .. DDS.KeyedString_Seq.Get_Length(received_data'Access)
	loop
	printArticle
	(DDS.KeyedString_Seq.Get (received_data'Access, i),
	DDS.SampleInfo_Seq.Get (sample_info'Access, i));
	end loop;
	reader.Return_Loan (received_data'Access, sample_info'Access);
	exception
	when DDS.NO_DATA =>
	null; -- ignore this error

--nothing to read; go back to sleep when DDS.ERROR =>

--an error occurred: stop reading raise;

end;

The read method takes several arguments:

❏Data and SampleInfo sequences: The first two arguments to read or take are lists: for the samples themselves and, in parallel, for meta-data about those samples. In most cases, you will pass these collections into the middleware empty, and the middleware will fill them for you. The objects it places into these sequences are loaned directly from the DataReader’s cache in order to avoid unnecessary object allocations or copies. When you are done with them, call return_loan.

If you would rather read the samples into your own memory instead of taking a loan from the middleware, simple pass in sequences in which the contents have already been deeply allocated. The DataReader will copy over the state of the objects you provide to it instead of loaning you its own objects. In this case, you do not need to call return_loan.

❏Number of samples to read: If you are only prepared to read a certain number of samples at a time, you can indicate that to the DataReader. In most cases, you will probably just use the constant LENGTH_UNLIMITED, which indicates that you are prepared to han- dle as many samples as the DataReader has available.

❏Sample state: The sample state indicates whether or not an individual sample has been observed by a previous call to read. (The News example uses this state to decide whether or not to append “(cached)” to the data it prints out.) By passing a sample state mask to

1.Ada support requires a separate add-on product, Ada 2005 Language Support.

5-14

Receiving Notifications When Data Delivery Is Late

the DataReader, you indicate whether you are interested in all samples, only those sam- ples you’ve never seen before, or only those samples you have seen before. The most com- mon value passed here is ANY_SAMPLE_STATE.

❏View state: The view state indicates whether the instance to which a sample belongs has been observed by a previous call to read or take. By passing a view state mask to the DataReader, you can indicate whether you’re interested in all instances, only those instances that you have never seen before (NEW instances), or only those instances that you have seen before (NOT_NEW instances). The most common value passed here is

ANY_VIEW_STATE.

❏Instance state: The instance state indicates whether the instance to which a sample belongs is still alive (ALIVE), whether it has been disposed (NOT_ALIVE_DISPOSED), or whether the DataWriter has simply gone away or stopped writing it without disposing it (NOT_ALIVE_NO_WRITERS). The most common value passed here is ANY_INSTANCE_STATE. For more information about the data lifecycle, consult the

RTI Core Libraries and Utilities User’s Manual and the API Reference HTML documenta- tion.

Unlike reading from a socket directly, or calling receive in JMS, a read or take is non-blocking: the call returns immediately. If no data was available to be read, it will return (in C and C++) or throw (in Java, .NET1, and Ada2) a NO_DATA result.

Connext also offers additional variations on the read() and take() methods to allow you to view different “slices” of your data at a time:

❏You can view one instance at a time with read_instance(), take_instance(), read_next_instance(), and take_next_instance().

❏You can view a single sample at a time with read_next_sample() and take_next_sample().

For more information about these and other variations, see the API Reference HTML documen- tation: open ReadMe.html, select a language, and look under Modules, Subscription Module, DataReader Support, FooDataReader. For Ada, open doc/html/api_ada/index.html and look under Subscription Module, DDSReaderModule, DDS.Typed_DataReader_Generic.

5.4.2Running & Verifying

To see the difference between read() and take() semantics, replace “read” with “take” (the argu- ments are the same), rebuild the example, and run again. You will see that “(cached)” is never printed, because every sample will be removed from the cache as soon as it is viewed for the first time.

5.5Receiving Notifications When Data Delivery Is Late

Many applications expect data to be sent and received periodically (or quasi-periodically). They typically expect at least one data sample to arrive during each period; a failure of data to arrive may or may not indicate a serious problem, but is probably something about which the applica- tion would like to receive notifications. For example:

❏A vehicle may report its current position to its home base every second.

1.RTI Connext .NET language binding is currently supported for C# and C++/CLI.

2.Ada support requires a separate add-on product, Ada 2005 Language Support.

5-15

Receiving Notifications When Data Delivery Is Late

❏Each sensor in a sensor network reports a new reading every 0.5 seconds.

❏A radar reports the position of each object that it’s tracking every 0.2 seconds.

If any vehicle, any sensor, or any radar track fails to yield an update within its promised period, another part of the system, or perhaps its human operator, may need to take a corrective action.

(In addition to built-in deadline support, Connext has other features useful to applications that publish and/or receive data periodically. For example, it’s possible to down-sample high-rate periodic data; see Subscribing Only to Relevant Data (Section 5.2).

5.5.1Implementation

Deadline enforcement is comprised of two parts: (1) QoS policies that specify the deadline con- tracts and (2) listener callbacks that are notified if those contracts are violated.

Deadlines are enforced independently for DataWriters and DataReaders. However, the Deadline QoS policy, like some other policies, has request-offer semantics: the DataWriter must offer a level of service that is the same or better than the level of service the DataReader requests. For example, if a DataWriter promises to publish data at least once every second, it will be able to communi- cate with a DataReader that expects to receive data at least once every two seconds. However, if the DataReader declares that it expects data twice a second, and the DataWriter only promises to publish updates only once a second, they will not be able to communicate.

5.5.1.1Offered Deadlines

A DataWriter promises to publish data at a certain rate by providing a finite value for the Dead-

line QoS policy, either in source code or in one or more XML configuration files. (Both mecha- nisms are functionally equivalent; the News example in C, C++, Java, and Ada1 uses XML files. For more information about this mechanism, see the chapter on “Configuring QoS with XML” in the RTI Core Libraries and Utilities User’s Manual.)

The file USER_QOS_PROFILES.xml in the News example for C, C++, Java, and Ada contains the following Deadline QoS policy configuration, which applies to both DataWriters and

DataReaders:

<deadline>

<period>

<sec>2</sec>

<nanosec>0</nanosec>

</period>

</deadline>

The DataWriter thus promises to publish at least one data sample—of each instance—every two seconds.

If a period of two seconds elapses from the time the DataWriter last sent a sample of a particular instance, the writer’s listener—if one is installed—will receive a callback to its on_offered_deadline_missed method. The News example does not actually install a DataWrit- erListener. See the section on requested deadlines below; the DataWriterListener works in a way that’s parallel to the DataReaderListener.

5.5.1.2Requested Deadlines

The DataReader declares that it expects to receive at least one data sample of each instance within a given period using the Deadline QoS policy. See the example XML configuration above.

1.Ada support requires a separate add-on product, Ada 2005 Language Support.

5-16

Receiving Notifications When Data Delivery Is Late

If the declared deadline period elapses since the DataReader received the last sample of some instance, that reader’s listener—if any—will be invoked. The listener’s on_requested_deadline_missed() will receive a call informing the application of the missed deadline.

To install a DataReaderListener in C++:

DDSDataReader* reader = participant->create_datareader( topic, DDS_DATAREADER_QOS_DEFAULT,

&_listener,	//	listener
DDS_STATUS_MASK_ALL);	//	all callbacks

if (reader == NULL) {

throw std::runtime_error("Unable to create DataReader");

}

In Java, it looks like this:

DataReader reader = participant.create_datareader( topic, Subscriber.DATAREADER_QOS_DEFAULT,

new ArticleDeliveryStatusListener(),	// listener
StatusKind.STATUS_MASK_ALL);	// all callbacks
if (reader == null) {

throw new IllegalStateException("Unable to create DataReader");

}

And in Ada1:

declare

readerListener : ArticleDeliveryStatusListener.Ref_Access; begin

reader := participant.Create_DataReader (topic.As_TopicDescription, DDS.Subscriber.DATAREADER_QOS_DEFAULT, readerListener'Unchecked_Access, -- listener DDS.STATUS_MASK_ALL); -- all callbacks

if reader = null then

Put_Line (Standard_Error, "Unable to create DataReader"); return;

end if;

end;

There are two listener-related arguments to provide:

❏Listener: The listener object itself, which must implement some subset of the callbacks defined by the DataReaderListener supertype.

❏Listener mask: Which of the callbacks you’d like to receive. In most cases, you will use one of the constants STATUS_MASK_ALL (if you are providing a non-null listener object) of STATUS_MASK_NONE (if you are not providing a listener). There are some cases in which you might want to specify a different listener mask; see the RTI Core Libraries and Utilities User’s Manual and API Reference HTML documentation for more information.

Let’s look at a very simple implementation of the on_requested_deadline_missed callback that prints the value of the key (i.e., the news outlet name) for the instance whose deadline was missed. You can see this in the News example provided with C, C++ and Java, and Ada.

1. Ada support requires a separate add-on product, Ada 2005 Language Support.

5-17

Receiving Notifications When Data Delivery Is Late

In C++, the code looks like this:

void ArticleDeliveryStatusListener::on_requested_deadline_missed( DDSDataReader* reader,

const DDS_RequestedDeadlineMissedStatus& status) { DDS_KeyedString keyHolder;

DDSKeyedStringDataReader* typedReader = DDSKeyedStringDataReader::narrow(reader);

typedReader->get_key_value(keyHolder, status.last_instance_handle);

std::cout << "->Callback: requested deadline missed: "

<<keyHolder.key

<<std::endl;

}

In Java:

public void on_requested_deadline_missed( DataReader reader,

RequestedDeadlineMissedStatus status) { KeyedString keyHolder = new KeyedString(); reader.get_key_value_untyped(keyHolder,

status.last_instance_handle); System.out.println("->Callback: requested deadline missed: " +

keyHolder.key);

}

And in Ada:

procedure On_Requested_Deadline_Missed

(Self : not null access Ref;

The_Reader : in DDS.DataReader.Ref'Class; Status : in DDS.RequestedDeadlineMissedStatus)

is

pragma Unreferenced (Self); pragma Unreferenced (The_Reader); pragma Unreferenced (Status);

begin

Put_Line ("->Callback: requested deadline missed."); end On_Requested_Deadline_Missed;

5.5.2Running & Verifying

Modify the file USER_QOS_PROFILES.xml to decrease the deadline to one second:

<deadline>

<period>

<sec>1</sec>

<nanosec>0</nanosec>

</period>

</deadline>

Note that, if you have a DataReader- or DataWriter-level deadline specified (inside the file's datareader_qos or datawriter_qos elements, respectively)—possibly because you previously modified the configuration in Section 5.2.3—it is overriding the topic-level configuration. Be careful that you don't modify a deadline specification that will only be overridden later and not take effect.

You will see output similar to Figure 5.5:

5-18

Receiving Notifications When Data Delivery Is Late

Figure 5.5 Using a Shorter Deadline

5-19

Chapter 6 Design Patterns for High Performance

In this chapter, you will learn how to implement some common performance-oriented design patterns. As you have learned, one of the advantages to using Connext is that you can easily tune your application without changing its code, simply by updating the XML-based Quality of Ser- vice (QoS) parameters.

We will build on the examples used in Building and Running “Hello, World” (Section 3.1) to demonstrate the different use-cases. The example applications (Hello_builtin, Hello_idl, and Hello_dynamic1), provide the same functionality but use different data types in order to help you understand the different type-definition mechanisms offered by Connext and their tradeoffs. They implement a simple throughput test: the publisher sends a payload to the subscriber, which periodically prints out some basic statistics. You can use this simple test to quickly see the effects of your system design choices: programming language, target machines, QoS configura- tions, and so on.

The QoS parameters do not depend on the language used for your application, and seldom on the operating system (there are few key exceptions), so you should be able to use the XML files with the example in the language of your choice.

This chapter describes:

❏Building and Running the Code Examples

❏Reliable Messaging

❏High Throughput for Streaming Data

❏Streaming Data over Unreliable Network Connections

6.1Building and Running the Code Examples

You can find the source for the Hello_builtin example for Java in ${NDDSHOME}/example/ JAVA/Hello_builtin; the equivalent source code in other supported languages is in the directo- ries ${NDDSHOME}/example/<language>. The Hello_idl and Hello_dynamic examples are in parallel directories under ${NDDSHOME}/example/<language>.

The examples perform these steps:

1.Parse their command-line arguments.

1.Dynamic types are not supported when using Ada 2005 Language Support.

6-1

Building and Running the Code Examples

Figure 6.1 Example Output from the Subscribing Application in C, C++, C# or Java

2. Check if the Quality of Service (QoS) file can be located.

The XML file that sets the Quality of Service (QoS) parameters is either loaded from the file USER_QOS_PROFILES.xml in the current directory of the program, or from the environment variable NDDS_QOS_PROFILES. For more information on how to use QoS profiles, see the chapter on "Configuring QoS with XML" in the RTI Core Libraries and Utilities User's Manual.

3.On the publishing side, send text strings as fast as possible, prefixing each one with a serial number.

4.On the subscribing side, receive these strings, keeping track of the highest number seen, as well as other statistics. Print these out periodically.

The steps for compiling and running the program are the same as mentioned in Building and Running “Hello, World” (Section 3.1).

Run the publisher and subscriber from the ${NDDSHOME}/example/<language>/ Hello_builtin directory using one of the QoS profile files provided in ${NDDSHOME}/exam- ple/QoS by copying it into your current working directory with the file name USER_QOS_PROFILES.xml. You should see output like the following from the subscribing application:

Understanding the Performance Results

You will see several columns in the subscriber-side output, similar to Figure 6.1.

❏Seconds from start: The number of seconds the subscribing application has been running. It will print out one line per second.

❏Total samples: The number of data samples that the subscribing application has received since it started running.

6-2

Reliable Messaging

❏Total lost samples: The number of samples that were lost in transit and could not be re- paired, since the subscribing application started running. If you are using a QoS profile configured for strict reliability, you can expect this column to always display 0. If you are running in best-effort mode, or in a limited-reliability mode (e.g., you have configured your History QoS policy to only keep the most recent sample), you may see non-zero val- ues here.

❏Current lost samples: The number of samples that were lost in transit and could not be repaired, since the last status line was printed. See the description of the previous col- umn for information about what values to expect here.

❏Average samples per second: The mean number of data samples received per second since the subscribing application started running. By default, these example applications send data samples that are 1 KB in size. (You can change this by passing a different size, in bytes, to the publishing application with --size.)

❏Current samples per second: The number of data samples received since the subscribing application printed the last status line.

❏Throughput megabits per second: The throughput from the publishing application to the subscribing application, in bits per second, since the subscribing application printed the previous status line. The value in this column is equivalent to the current samples per second multiplied by the number of bits in an individual sample.

With small sample sizes, the fixed "cost" of traversing your operating system's network stack is greater than the cost of actually transmitting the data; as you increase the sample size, you will see the throughput more closely approach the theoretical throughput of your network.

By batching multiple data samples into a single network packet, as the high-throughput example QoS profile does, you should be able to saturate a gigabit Ethernet network with samples sizes as small as 100-200 bytes. Without batching samples, you should be able to saturate the network with samples of a few kilobytes. The difference is due to the perfor- mance limitations of the network transport; enterprise-class platforms with commodity Ethernet interfaces can typically execute tens of thousands of send()’s per second. In con- trast, saturating a high-throughput network link with data sizes of less than a kilobyte requires hundreds of thousands of samples per second.

Is this the best possible performance?

The performance of an application depends heavily on the operating system, the network, and how it configures and uses the middleware. This example is just a starting point; tuning is important for a production application. RTI can help you get the most out of your platform and the middleware.

To get a sense for how an application's behavior changes with different QoS contracts, try the other provided example QoS profiles and see how the printed results change.

6.2Reliable Messaging

Packets sent by a middleware may be lost by the physical network or dropped by routers, switches and even the operating system of the subscribing applications when buffers become full. In reliable messaging, the middleware keeps track of whether or not data sent has been received by subscribing applications, and will resend data that was lost on transmission.

Like most reliable protocols (including TCP), the reliability protocol used by RTI uses additional packets on the network, called metadata, to know when user data packets are lost and need to

6-3

Reliable Messaging

be resent. RTI offers the user a comprehensive set of tunable parameters that control how many and how often metadata packets are sent, how much memory is used for internal buffers that help overcome intermittent data losses, and how to detect and respond to a reliable subscriber that either falls behind or otherwise disconnects.

When users want applications to exchange messages reliably, there is always a need to trade-off between performance and memory requirements. When strictly reliable communication is enabled, every written sample will be kept by Connext inside an internal buffer until all known reliable subscribers acknowledge receiving the sample1.

If the publisher writes samples faster than subscribers can acknowledge receiving, this internal buffer will eventually be completely filled, exhausting all the available space—in that case, fur- ther writes by the publishing application will block. Similarly, on the subscriber side, when a sample is received, it is stored inside an internal receive buffer, waiting for the application to take the data for processing. If the subscribing application doesn't take the received samples fast enough, the internal receive buffer may fill up—in that case, newly received data will be dis- carded and would need to be repaired by the reliable protocol.

Although the size of those buffers can be controlled from the QoS, you can also use QoS to con- trol what Connext will do when the space available in one of those buffers is exhausted. There are two possible scenarios for both the publisher and subscriber:

Publishing side: If write() is called and there is no more room in the DataWriter’s buffer, Connext can:

1.Temporarily block the write operation until there is room on this buffer (for example, when one or more samples is acknowledged to have been received from all the subscrib- ers).

2.Drop the oldest sample from the queue to make room for the new one.

Subscribing side: If a sample is received (from a publisher) and there is no more room on the

DataReader’s buffer:

1.Drop the sample as if it was never received. The subscribing application will send a neg- ative acknowledgement requesting that the sample be resent.

2.Drop the oldest sample from the queue to make room for the new one.

6.2.1Implementation

There are many variables to consider, and finding the optimum values to the queue size and the right policy for the buffers depends on the type of data being exchanged, the rate of which the data is written, the nature of the communication between nodes and various other factors.

The RTI Core Libraries and Utilities User’s Manual dedicates an entire chapter to the reliability pro- tocol, providing details on choosing the correct values for the QoS based on the system configu- ration. For more information, refer to Chapter 10 in the User’s Manual.

The following sections highlight the key QoS settings needed to achieve strict reliability. In the reliable.xml QoS profile file, you will find many other settings besides the ones described here. A detailed description of these QoS is outside the scope of this document, and for further infor- mation, refers to the comments in the QoS profile and in the User’s Manual.

1. Connext also supports reliability based only on negative acknowledgements ("NACK-only reliability"). This fea- ture is described in detail in the User's Manual (Section 6.5.2.3) but is beyond the scope of this document.

6-4

Reliable Messaging

6.2.1.1Enable Reliable Communication

The QoS that control the kind of communication is the Reliability QoS of the DataWriter and

DataReader:

<datawriter_qos>

...

<reliability> <kind>RELIABLE_RELIABILITY_QOS</kind> <max_blocking_time>

<sec>5</sec>

<nanosec>0</nanosec> </max_blocking_time>

</reliability>

...

</datawriter_qos>

...

<datareader_qos> <reliability>

<kind>RELIABLE_RELIABILITY_QOS</kind> </reliability>

</datareader_qos>

This section of the QoS file enables reliability on the DataReader and DataWriter, and tells the middleware that a call to write() may block up to 5 seconds if the DataWriter’s cache is full of unacknowledged samples. If no space opens up in 5 seconds, write() will return with a timeout indicating that the write operation failed and that the data was not sent.

6.2.1.2Set History To KEEP_ALL

The History QoS determines the behavior of a DataWriter or DataReader when its internal buffer fills up. There are two kinds:

❏KEEP_ALL: The middleware will attempt to keep all the samples until they are acknowl- edged (when the DataWriter’s History is KEEP_ALL), or taken by the application (when the DataReader’s History is KEEP_ALL).

❏KEEP_LAST: The middleware will discard the oldest samples to make room for new sam- ples. When the DataWriter’s History is KEEP_LAST, samples are discarded when a new call to write() is performed. When the DataReader’s History is KEEP_LAST, samples in the receive buffer are discarded when new samples are received. This kind of history is associated with a depth that indicates how many historical samples to retain.

<datawriter_qos> <history>

<kind>KEEP_ALL_HISTORY_QOS</kind> </history>

...

</datawriter_qos>

...

<datareader_qos> <history>

<kind>KEEP_ALL_HISTORY_QOS</kind> </history>

...

</datareader_qos>

The above section of the QoS profile tells RTI to use the policy KEEP_ALL for both DataReader and DataWriter.

6-5

High Throughput for Streaming Data

6.2.1.3Controlling Middleware Resources

With the ResourceLimits QosPolicy, you have full control over the amount of memory used by the middleware. In the example below, we specify that both the reader and writer will store up to 10 samples (if you use a History kind of KEEP_LAST, the values specified here must be con-

sistent with the value specified in the History’s depth).

<datawriter_qos> <resource_limits>

<max_samples>10</max_samples> </resource_limits>

...

</datawriter_qos>

...

<datareader_qos> <resource_limits>

<max_samples>2</max_samples> </resource_limits>

...

</datareader_qos>

The above section tells RTI to allocate a buffer of 10 samples for the DataWriter and 2 for the DataReader. If you do not specify any value for max_samples, the default behavior is for the mid- dleware to allocate as much space as it needs.

One important function of the Resource Limits policy, when used in conjunction with the Reli- ability and History policies, is to govern how far "ahead" of its DataReaders a DataWriter may get before it will block, waiting for them to catch up. In many systems, consuming applications can- not acknowledge data as fast as its producing applications can put new data on the network. In such cases, the Resource Limits policy provides a throttling mechanism that governs how many sent-but-not-yet-acknowledged samples a DataWriter will maintain. If a DataWriter is configured for reliable KEEP_ALL operation, and it exceeds max_samples, calls to write() will block until the writer receives acknowledgements that will allow it to reclaim that memory.

If you see that your reliable publishing application is using an unacceptable amount of memory, you can specify a finite value for max_samples. By doing this, you restrain the size of the DataWriter's cache, causing it to use less memory; however, a smaller cache will fill more quickly, potentially causing the writer to block for a time when sending, decreasing throughput. If decreased throughput proves to be an issue, you can tune the reliability protocol to process acknowledgements and repairs more aggressively, allowing the writer to clear its cache more effectively. A full discussion of the relevant reliability protocol parameters is beyond the scope of this example. However, you can find a useful example in high_throughput.xml. Also see the documentation for the DataReaderProtocol and DataWriterProtocol QoS policies in the on-line API documentation.

6.3High Throughput for Streaming Data

This design pattern is useful for systems that produce a large number of small messages at a high rate.

In such cases, there is a small but measurable overhead in sending (and in the case of reliable communication, acknowledging) each message separately on the network. It is more efficient for the system to manage many samples together as a group (referred to in the API as a batch) and then send the entire group in a single network packet. This allows Connext to minimize the over- head of building a datagram and traversing the network stack.

6-6

High Throughput for Streaming Data

Batching increases throughput when writing small samples at a high rate. As seen in Figure 6.2, throughput can be increased several-fold, much more closely approaching the physical limita- tions of the underlying network transport.

Figure 6.2 Benefits of Batching

Batching delivers tremendous benefits for messages of small size.

Figure 6.3 Benefits of Batching: Sample Rates

A subset of the batched throughput data above, expressed in terms of samples per second.

Collecting samples into a batch implies that they are not sent on the network (flushed) immedi- ately when the application writes them; this can potentially increase latency. However, if the application sends data faster than the network can support, an increased share of the network's available bandwidth will be spent on acknowledgements and resending dropped data. In this case, reducing that meta-data overhead by turning on batching could decrease latency even while increasing throughput. Only an evaluation of your system's requirements and a measurement of its actual performance will indicate whether batching is appropriate. Fortunately, it is easy to enable and tune batching, as you will see below.

Batching is particularly useful when the system has to send a large number of small messages at a fast rate. Without this feature enabled, you may observe that your maximum throughput is

6-7

High Throughput for Streaming Data

less than the maximum bandwidth of your network. Simultaneously, you may observe high CPU loads. In this situation, the bottleneck in your system is the ability of the CPU to send pack- ets through the OS network stack.

For example, in some algorithmic trading applications, market data updates arrive at a rate of from tens of thousands of messages per second to over a million; each update is some hundreds of bytes in size. It is often better to send these updates in batches than to publish them individu- ally. Batching is also useful when sending a large number of small samples over a connection where the bandwidth is severely constrained.

6.3.1Implementation

RTI can automatically flush batches based on the maximum number of samples, the total batch size, or elapsed time since the first sample was placed in the batch, whichever comes first. Your application can also flush the current batch manually. Batching is completely transparent on the subscribing side; no special configuration is necessary.

Figure 6.4 Batching Implementation

RTI collects samples in a batch until the batch is flushed.

For more information on batching, see the RTI Core Libraries and Utilities User’s Manual (Section 6.5.1) or API Reference HTML documentation (the Batch QosPolicy is described in the Infra- structure Module).

Using the batching feature is simple—just modify the QoS in the publishing application’s con- figuration file.

For example, to enable batching with a batch size of 100 samples, set the following QoS in your XML configuration file:

<datawriter_qos>

...

<batch>

<enable>true</enable> <max_samples>100</max_samples>

</batch>

...

</datawriter_qos>

To enable batching with a maximum batch size of 8K bytes:

<datawriter_qos>

...

<batch>

<enable>true</enable> <max_data_bytes>8192</max_data_bytes>

</batch>

...

</datawriter_qos>

6-8

High Throughput for Streaming Data

To force the DataWriter to send whatever data is currently stored in a batch, use the DataWriter’s flush() operation.

6.3.2Running & Verifying

1.To run and verify the new QoS parameters, first run Hello_builtin using a simple reli- able configuration—batching is turned off. Copy reliable.xml from $NDDSHOME/ example/QoS to $NDDSHOME/example/<language>/Hello_builtin and rename it to USER_QOS_PROFILES.xml. Observe the results.

Figure 6.5 Initial Performance Results

2.Next, run the example again with a configuration that includes batching. Copy high_throughput.xml from ${NDDSHOME}/example/QoS to ${NDDSHOME}/ example/<language>/Hello_builtin and rename it to USER_QOS_PROFILES.xml.

3.Run Hello_builtin again. Verify that the throughput numbers have improved.

Figure 6.6 Improved Performance Results

6-9

Streaming Data over Unreliable Network Connections

6.4Streaming Data over Unreliable Network Connections

Systems face unique challenges when sending data over lossy networks that also have high- latency and low-bandwidth constraints—for example, satellite and long-range radio links. While sending data over such a connection, a middleware tuned for a high-speed, dedicated Gigabit connection would throttle unexpectedly, cause unwanted timeouts and retransmissions, and ultimately suffer severe performance degradation.

For example, the transmission delay in satellite connections can be as much as 500 milliseconds to 1 second, which makes such a connection unsuitable for applications or middleware tuned for low-latency, real-time behavior. In addition, satellite links typically have lower bandwidth, with near-symmetric connection throughput of around 250–500 Kb/s and an advertised loss of approximately 3% of network packets. (Of course, the throughput numbers will vary based on the modem and the satellite service.) In light of these facts, a distributed application needs to tune the middleware differently when sending data over such networks.

Connext is capable of maintaining liveliness and application-level QoS even in the presence of sporadic connectivity and packet loss at the transport level, an important benefit in mobile, or otherwise unreliable networks. It accomplishes this by implementing a reliable protocol that not only sequences and acknowledges application-level messages, but also monitors the liveliness of the link. Perhaps most importantly, it allows your application to fine-tune the behavior of this protocol to match the characteristics of your network. Without this latter capability, communica- tion parameters optimized for more performant networks could cause communication to break down or experience unacceptable blocking times, a common problem in TCP-based solutions.

6.4.1Implementation

When designing a system that demands reliability over a network that is lossy and has high latency and low throughput, it is critical to consider:

❏How much data you send at one time (e.g., your sample or batch size).

❏How often you send it.

❏The tuning of the reliability protocol for managing meta- and repair messages.

It is also important to be aware of whether your network supports multicast communication; if it does not, you may want to explicitly disable it in your middleware configuration (e.g., by using the NDDS_DISCOVERY_PEERS environment variable or setting the initial_peers and multicast_receive_address in your Discovery QoS policy; see the API Reference HTML docu- mentation).

6.4.1.1Managing Your Sample Size

Pay attention to your packet sizes to minimize or avoid IP-level fragmentation. Fragmentation can lead to additional repair meta-traffic that competes with the user traffic for bandwidth. Ethernet-like networks typically have a frame size of 1500 bytes; on such networks, sample sizes (or sample fragment sizes, if you've configured Connext to fragment your samples) should be kept to approximately 1400 bytes or less. Other network types will have different fragmentation thresholds.

The exact size of the sample on the wire will depend not only on the size of your data fields, but also on the amount of padding introduced to ensure alignment while serializing the data.

6-10

Streaming Data over Unreliable Network Connections

Figure 6.7 shows how an application's effective throughput (as a percentage of the theoretical capacity of the link) increases as the amount of data in each network packet increases. To put this relationship in another way: when transmitting a packet is expensive, it's advantageous to put as much data into it as possible. However, the trend reverses itself when the packet size becomes larger than the maximum transmission unit (MTU) of the physical network.

Figure 6.7 Example Throughput Results over VSat Connection

Correlation between sample size and bandwidth usage for a satellite connection with 3% packet loss ratio.

To understand why this occurs, remember that data is sent and received at the granularity of application samples but dropped at the level of transport packets. For example, an IP datagram 10 KB in size must be fragmented into seven (1500-byte) Ethernet frames and then reassembled on the receiving end; the loss of any one of these frames will make reassembly impossible, lead- ing to an effective loss, not of 1500 bytes, but of over 10 thousand bytes.

On an enterprise-class network, or even over the Internet, loss rates are very low, and therefore these losses are manageable. However, when loss rates reach several percent, the risk of losing at least one fragment in a large IP datagram becomes very large1. Over an unreliable protocol like UDP, such losses will eventually lead to near-total data loss as data size increases. Over a proto- col like TCP, which provides reliability at the level of whole IP datagrams (not fragments), mounting losses will eventually lead to the network filling up with repairs, which will them- selves be lost; the result can once again be near-total data loss.

To solve this problem, you need to repair data at the granularity at which it was lost: you need, not message-level reliability, but fragment-level reliability. This is an important feature of Con- next. When sending packets larger than the MTU of your underlying link, use RTI's data frag-

1. Suppose that a physical network delivers a 1 KB frame successfully 97% of the time. Now suppose that an appli- cation sends a 64 KB datagram. The likelihood that all fragments will arrive at their destination is 97% to the 64th power, or less than 15%.

6-11

Streaming Data over Unreliable Network Connections

mentation and asynchronous publishing features to perform the fragmentation at the middleware level, hence relieving the IP layer of that responsibility.

<datawriter_qos>

...

<publish_mode> <kind>ASYNCHRONOUS_PUBLISH_MODE_QOS</kind> <flow_controller_name>

DDS_DEFAULT_FLOW_CONTROLLER_NAME </flow_controller_name>

</publish_mode>

...

</datawriter_qos>

<participant_qos>

...

<property>

<value>

<element>

<name> dds.transport.UDPv4.builtin.parent.message_size_max

</name>

<value>1500</value>

</element>

</value>

</property>

...

</participant_qos>

The DomainParticipant's dds.transport.UDPv4.builtin.parent.message_size_max property sets the maximum size of a datagram that will be sent by the UDP/IPv4 transport. (If your applica- tion interfaces to your network over a transport other than UDP/IPv4, the name of this property will be different.) In this case, it is limiting all datagrams to the MTU of the link (assumed, for the sake of this example, to be equal to the MTU of Ethernet).

At the same time, the DataWriter is configured to send its samples on the network, not synchro- nously when write() is called, but in a middleware thread. This thread will "flow" datagrams onto the network at a rate determined by the FlowController1 identified by the flow_controller_name. In this case, the FlowController is a built-in instance that allows all data to be sent immediately. In a real-world application, you may want to use a custom FlowCon- troller that you create and configure in your application code. Further information on this topic is beyond the scope of this example. For more information on asynchronous publishing, see Sec- tion 6.4.1 in the RTI Core Libraries and Utilities User's Manual. You can also find code examples demonstrating these capabilities online in the Solutions area of the RTI Support Portal, accessi- ble from https://support.rti.com/. Navigate to Code Examples and search for Asynchronous Publication.

1. FlowControllers are not supported when using Ada 2005 Language Support.

6-12

Streaming Data over Unreliable Network Connections

6.4.1.2Acknowledge and Repair Efficiently

Piggyback heartbeat with each sample. A DataWriter sends "heartbeats"—meta-data messages announcing available data and requesting acknowledgement—in two ways: periodically and "piggybacked" into application data packets. Piggybacking heartbeats aggressively ensures that the middleware will detect packet losses early, while allowing you to limit the number of extra- neous network sends related to periodic heartbeats.

<datawriter_qos>

...

<resource_limits>

<!-- Used to configure piggybacks w/o batching --> <max_samples>

20 <!-- An arbitrary finite size --> </max_samples>

</resource_limits> <writer_resource_limits>

<!-- Used to configure piggybacks w/ batching; see below --> <max_batches>

20 <!-- An arbitrary finite size --> </max_batches>

</writer_resource_limits> <protocol>

<rtps_reliable_writer> <heartbeats_per_max_samples>

20 <!-- Set same as max_samples --> </heartbeats_per_max_samples>

<rtps_reliable_writer> </protocol>

...

</datawriter_qos>

The heartbeats_per_max_samples parameter controls how often the middleware will piggy- back a heartbeat onto a data message: if the middleware is configured to cache 10 samples, for example, and heartbeats_per_max_samples is set to 5, a heartbeat will be piggybacked unto every other sample. If heartbeats_per_max_samples is set equal to max_samples, this means that a heartbeat will be sent with each sample.

6.4.1.3Make Sure Repair Packets Don’t Exceed Bandwidth Limitation

Applications can configure the maximum amount of data that a DataWriter will resend at a time using the max_bytes_per_nack_response parameter. For example, if a DataReader sends a nega- tive acknowledgement (NACK) indicating that it missed 20 samples, each 10 KB in size, and max_bytes_per_nack_response is set to 100 KB, the DataWriter will only send the first 10 sam- ples. The DataReader will have to NACK again to receive the remaining 10 samples.

6-13

Streaming Data over Unreliable Network Connections

In the following example, we limit the number of bytes so that we will never send more data than a 256 Kb/s, 1-ms latency link can handle over one second:

<datawriter_qos>

...

<protocol> <rtps_reliable_writer>

<max_bytes_per_nack_response> 28000

</max_bytes_per_nack_response> </rtps_reliable_writer>

</protocol>

...

</datawriter_qos>

6.4.1.4Use Batching to Maximize Throughput for Small Samples

If your application is sending data continuously, consider batching small samples to decrease the per-sample overhead. Be careful not to set your batch size larger than your link’s MTU; see Managing Your Sample Size (Section 6.4.1.1).

For more information on how to configure throughput for small samples, see High Throughput for Streaming Data (Section 6.3).

6-14