[Bug] Performance drop routing 1kHz publications since 0.11.0

[franz6ko]

Describe the bug

This is related to this other bug in zenoh_plugin_ros2dds here.

The solution for me was the same, rolling back to a previous version. In this case 0.10.0-rc-1

To reproduce

Updated to last version.
Check rate of high speed messages (previously 1000Hz dropped to 200Hz)

System info

Ubuntu 20.04

[JEnoch]

Thanks for this feedback.
Could you please detail the scenario in which you see such performance drop ?
Is it some kind of throughput test ? With which size of payload ?

[franz6ko]

Hi @JEnoch,

Yes, the test was the same I explain in this other issue but in this case between the release 1.0.0-alpha.4 and the 0.10.0-rc-1.

In this case, a 1000Hz topic will drop to around 200Hz on 1.0.0-alpha.4 but would stay at 1000Hz on 0.10.0-rc-1.

I'm using it with ROS2 Foxy on Ubuntu 20.04. The topic message is basic structure holding some floats.

[JEnoch]

Probably the same cause than explained in eclipse-zenoh/zenoh-plugin-ros2dds#192 (comment) .
The issue is caused by some weird interaction between tokio and CycloneDDS.
Investigation in progress...

[JEnoch]

Zenoh has now been fixed (in branch dev/1.0.0) to adapt its batching mechanism to the tokio behaviour:
eclipse-zenoh/zenoh#1335
I can confirm that with default configuration (i.e. without configuring the express flag for publishers) the routing of 1kHz publications is no longer suffering performance drop.

This repo will be sync to use the fixed Zenoh version in branch dev/1.0.0 in the next few days.

[franz6ko]

Great to hear!! Thanks for your support. I'm happy to have contributed to this issue.