Github user revans2 commented on the issue:
https://github.com/apache/storm/pull/2203
I also want to make sure that we take the performance impact of this patch
into consideration. With the old metrics system we are sub-sampling the
metrics for performance reasons. With this patch we are not. When I run
ThroughputVsLatency, not perfect I know. With this patch the maximum
throughput I am able to get is about 48,000 sentences per second. Without this
patch I can do about 60,000 reliably. This is a 20% drop in maximum
throughput. Note that all I did was adjust the target throughput and set max
spout pending to 500.
---
