How to test latency performance with different data size?

[chaoli2]

I notice that there are multiple message types defined with different data size. I've tried "Array1k"(default size in README), "Array4k", "Array16k", "Array60k"(why noy 64k...), "Array1m" and "Array2m" with below command or similar:

ros2 run performance_test perf_test -c ROS2 -l log -t Array4k --max_runtime 120

Above message types all work well.

But when I move forward to "Array4m", I got below error message:

ros2 run performance_test perf_test -c ROS2 -l log -t Array4m --max_runtime 120

terminate called after throwing an instance of 'std::runtime_error'
  what(): a topic with the requested name does not exist or communication mean not supported.

When I switch my test to PointCloud message type, there are not any responding when executing below command

ros2 run performance_test perf_test -c ROS2 -l log -t PointCloud1m --max_runtime 120

What command and parameter should I use when testing PointCloud message type?

[deeplearningrobotics]

@chaoli2: I think there are two seperate problems here:

ros2 run performance_test perf_test -c ROS2 -l log -t Array4m --max_runtime 120

terminate called after throwing an instance of 'std::runtime_error'
  what(): a topic with the requested name does not exist or communication mean not supported.

This looks like a bug. Let me investigate.

2. Try to reduce the rate you send messages. Maybe your system is overloaded.

[chaoli2]

@deeplearningrobotics
Thanks for your reply. How can I reduce the rate? Is there any parameter to do this?

[deeplearningrobotics]

@chaoli2: Yes. Running ros2 run performance_test perf_test --help will give you all the parameters. You should try --rate 10 as a start which will publish 10 samples a second. The default I think is a 1000 samples.

[chaoli2]

@deeplearningrobotics
Thanks. The parameter works, but when I run

ros2 run performance_test perf_test -c ROS2 -l log -t PointCloud1m --max_runtime 120 --rate 10

I got a new error:

terminate called after throwing an instance of 'rclcpp::exceptions::RCLError'
  what(): could not create publisher: Liveliness QoS is not yet supported for fastrtps., at /tmp/binarydeb/rosdashing-rmw-fastrtps-shared-cpp-0.7.4/src/qos.cpp:167, at /tmp/binarydeb/ros-dashing-rcl-0.7.5/src/rcl/publisher.c:171

Do you get any idea about this?

[dejanpan]

@chaoli2 the error is what it indeed is - FastRTPS in Dashing namely indeed does not have Liveliness QoS supported.

@esteve @MiguelCompany do you know why is performance_test trying to use it?

[MiguelCompany]

@esteve @MiguelCompany do you know why is performance_test trying to use it?

I think this is the same as in #66. Please see my comment there for a possible solution.

[chaoli2]

@MiguelCompany @dejanpan
Thanks for your update. I got the PointCloud performance test results of "512K", "1m", "2m" and "4m", but failed when testing "PointCloud8m".

ros2 run performance_test perf_test -c ROS2 -l log -t PointCloud8m --max_runtime 120 --rate 10

terminate called after throwing an instance of 'std::runtime_error'
  what(): a topic with the requested name does not exist or communication mean not supported.