Github user jmartell7 commented on the issue: https://github.com/apache/storm/pull/2399 Performance didn't seem to be affected by adding the metric. Using ThroughputVsLatency on my MBP (2.3 GHz i7, 16GB RAM) with 35k tuples / sec, 4 workers, 4 splitters, 4 counters, 4 spouts, I saw the following metrics: Avg latency was 14.3ms with master and 14.0ms with the metric enabled, 99% latency was 17.0 and 16.9 respectively, and cpu was 3.05 and 3.09 respectively. Additional details are available for the asking.
---