[Discourse.ros.org] [Next Generation ROS] FastRTPS throughput issues

[Secretary Birds]

SecretaryBirds October 13

I am publishing some data at a very high frequency (1000 Hz). It's essentially a time server, so the data isn't huge. Just seconds and nanoseconds. This is my way of replicating the ROS Time that exists in ROS 1. I created another node that subscribes to this time data. I am hitting a roadblock using Alpha8 and FastRTPS, where I call spin_some(node) and it basically never comes back. My intuition is that it spends the whole time processing messages for the subscription, never gets ahead of the incoming data, so the spin_some can never return. Any experience with this? Or pointers? Is this something that's an inherent limitation of FastRTPS? I didn't notice this with Alpha7, but I was using OpenSplice. I assume the ROS layer didn't change enough to make this a ROS issue, but rather a FastRTPS issue. I can run that test to verify. Thanks.

---

Visit Topic or reply to this email to respond.

To unsubscribe from these emails, click here.

[Jaime Martin Losa]

Jaime_Martin_Losa October 14

Hi, 1000 Hz is not so high, and I think you should be able to do it with our Fast RTPS. We have throughput tests sending at higher rates with larger data. Without seeing the actual test you are running is hard to answer. Could you try to reproduce the problem on Fast RTPS and post an issue on our Github ? Thanks, Jaime Martin Losa CEO eProsima - The middleware experts Phone: + 34 607 91 37 45 Twitter: @JaimeMartinLosa www.eProsima.com

[William Woodall]

wjwwood October 17

It's hard to say why you're experiencing this behavior. If you can provide a simple example where spin_some doesn't return, we can have a look at it. I'd have to look into how spin_some is implemented, but I was aware of that pitfall when we were writing it. It could still be broken though.