Hello, I have a trident drpc topology. I am trying to reduce the complete latency which is around 10 seconds for now. I see that the bottleneck looks like the process latency which is 100x more than the execute latency. It is my assumption that we don’t do explicit ack in Trident so I would expect the ack been sent at the end of execute method.
To make more sense of high process latency to that of execute latency, 1. Do you believe high latency is caused by the executor_disruptor_send_queue/thread ? I did try with “topology.executor.send.buffer.size 16384” 2. What other metrics should I look for high process latency? Really appreciate if somebody can throw some light have been trying to resolve for a week now. Let me know if you I can provide more details. Regards, Mohan