Discovery protocol in DDS

6 posts / 0 new
Last post
Offline
Joined: 11/06/2013
Posts: 3
Discovery protocol in DDS

Hi,

I am going to do my master thesis in simple discovery protocol (SDP) to reduce the required overhead and improve its scalability. The current discovery protocol (SDP) uses flooding multicast storm in the network since that every Participant should sent to all other peers in GDS information about all the Endpoints that under that Participant. I want to propose to use a structured DHT overlay as an infrastructure for discovery protocol so it is going to replace the flooding method with sort of structured tree-based network using one of DHT algorithms.

What is the feasibility of such implementation for DDS and is there any negative expected outcomes for such approach?

Regards

Offline
Joined: 05/23/2013
Posts: 47

Hi,

The following paper gives you some ideas of expected outcomes using DHT overlay for discovery protocol even though they provided simulation results.
http://link.springer.com/chapter/10.1007%2F978-3-642-26010-0_15

Thanks,
Kyoungho

JavierPovedano's picture
Offline
Joined: 06/05/2013
Posts: 16

Hi Khalid,

Here are two more papers related with scalable discovery on DDS:

RELOAD extension for data discovery and transfer in data-centric publish–subscribe environments

http://www.sciencedirect.com/science/article/pii/S0920548913000743

Bloom filter-based discovery protocol for DDS middleware

http://www.sciencedirect.com/science/article/pii/S0743731511000876

Hope you find them useful for your research

 
Javier

Offline
Joined: 11/06/2013
Posts: 3

I appreciate your gentle help and comments that you provide in this topic.

kyoungho, I have read that paper a long ago and I never see any citation for it so far. I did not even see it mentioned in RTI community. So, I am so sure about it credibility. So, that motivates me to share the problem in RTI community. The simulation that they did is quit basic and it might not fulfill all OMG RTPS requirement. 

Javier Povedano, I have come through your second paper I was inspired by it. There is a recent enhancement for it by Putra, H.A. in a paper titled by: “Discovery protocol for data distribution service in naval warships using extended counting bloom filters” to eliminate false positive problem. I am more interested to minimize the extra load and stress in the network which caused by data dissemination that done in a mesh method and replace with structured DHT overlay broadcast that going to be used under SDP for the purpose of distributing discovery messages with a minimum network congestion and overhead.  For your first one, it seems so awesome but it’s not directly related to SDP which I am focusing in my current research.

Regards,

Khalid 

Offline
Joined: 11/06/2013
Posts: 3

Hi,
I need to understand the mechanism of message dissemination in simple discovery protocol (for both PDP and EDP). Does it consider as IP multicast or Overlay multicast?

Offline
Joined: 05/23/2013
Posts: 47

Hi Khalid,

It does not mean that the work is not credible if it is not mentioned in this community. However, I think it would be intresting to see the actual results instead of the simulation ones. Can you please specify that "it might not fulfill all OMG RTPS requirement"? I am not involed in that work, but I am just curous about what it means :)

I believe both SPDP and SEDP disseminate discovery messages over IP multicast and possibly unicast if you change some QoS settings. I do not know about discovery protocols over overlay multicast.