Storm / Trident / DRPC Expert to Troubleshoot Latency Issues

Project Description

We have a trident drpc topology with heavy parallelism to perform event based queries. Our drpc topolgy generally runs with a latency of 8 to 12 seconds whereas we'd expect the latency to be 1 - 2 seconds. 

During testing, we have observed that our execute latency for each function is under 10 to 15ms but the process latency is higher (in the order of 100s to 1000s of ms). The topology capacity is under 0.5. We've tried to improve the process latency by increasing the executor queue buffer size and adjusting other parameters without noticeable difference.  Attached is a screenshot of the topology running.

The expert must have Storm Trident expertise to help us scale and tune our Trident topology.