ros2 galactic interprocess performance
Hi, we did a lot of ipc performance measuring to figure out whether ROS2 + Cyclone DDS is full filling performance requirements for autonomous driving prototyping. I made some latency measurements with a simple pub/sub setup (see ros2_latency_ipc). Because most of our interfaces use dynamic sized messages, we can not really utilize iceoryx shm here. Is there any way "RMW Cyclone DDS only" to decrease latencies for large but also for small payload ? QOS, XML configuration ... ? Measuring results and platform information can be found here for Windows 10 and Ubuntu 20.04.
Not a definitive answer, but a point of data: At the university, we use ROS 2 for exactly this application. Out of the box, with default settings, it works just fine. The students enjoy it a lot. Additionally, ROS is much easier to start with than other frameworks we tried in the past.