This message was deleted.

That's a good question. We are long overdue on a benchmark. Maybe @Sean or @larme (shenyang) can comment in on this topic

Hi @Andrew Sica, in your experiment, what is your container setup (CPUs, memory)? Were you able to saturate the CPU at peak throughput? Could you please speak more about the overhead you saw during batching? Let’s focus more on the 1.0 version for now.

Hi @Sean , I'm going to rerun tomorrow/Monday and will gather some results in a dedicated env. Will post back