6.2.8. Performance and Scalability

6.2.8.1. [Major] Performance issues when using FlatData with payload encryption or compression

You may have seen performance issues when using the FlatData language binding along with compression or payload encryption. In this case, the number of copies of each sample was not reduced to two, as is expected when using FlatData. (See “34.1.4 FlatData Language Binding” in the Core Libraries User’s Manual.) This issue removed the performance improvement that FlatData provides, but only when compression or payload encryption was enabled. This problem did not occur when using FlatData without compression or payload encryption. This problem has been fixed.

[RTI Issue ID CORE-11262]

6.2.8.2. [Major] Transport utilization metrics overflowed in applications with high throughput *

Transport utilization periodic metrics (like dds_participant_udpv4_usage_in_net_bytes_count or dds_participant_udpv6_usage_out_net_bytes_count) could overflow in high-throughput applications (for example, applications that wrote and/or received large data with high frequency). If the polling period of Monitoring Library 2.0 (previously called Observability Library) was big enough, the variation of the metrics in the period of time did not fit into a 32-bit integer.

If a metric overflowed, an error message like the following was produced:

ERROR RTI_Monitoring_getTransportUtilizationStatistics:TYPE CONVERSION FAILURE | count (4421753352) exceeds max. representable UINT32 for metric with metricGroupIndex 22

The metric was not propagated to Observability Dashboards in this case.

To mitigate this issue, transport utilization count metrics have been promoted to a 64-bit integer. Reducing the polling period also makes the overflow less likely.

[RTI Issue ID MONITOR-597]

6.2.8.3. [Minor] Performance degradation when using FlatData with ContentFilteredTopics

In previous releases, a DataWriter using FlatData and doing writer-side filtering for DataReaders using ContentFilteredTopics may have done more data copies than necessary, leading to suboptimal performance. This problem has been fixed.

[RTI Issue ID CORE-13250]

* This bug does not affect you if you are upgrading from 6.1.x or earlier.