Github user franz1981 commented on the issue: https://github.com/apache/activemq-artemis/pull/1793 @RaiSaurabh @michaelandrepearce I've run a couple of benchmarks and the results are strange, perf wire the improvements are on par or (slightly) less than https://github.com/apache/activemq-artemis/pull/1849, but the number are "strange" eg: ``` 4 producers - 4 consumers ------------------------------------ EndToEnd Throughput: 145985 ops/sec EndToEnd SERVICE-TIME Latencies distribution in MICROSECONDS mean 10032775.17 min 9999220.74 50.00% 10066329.60 90.00% 10066329.60 99.00% 10066329.60 99.90% 10066329.60 99.99% 10066329.60 max 10066329.60 count 400000 --------------------------------- 1 producer - 1 consumer --------------------------------- EndToEnd Throughput: 78247 ops/sec EndToEnd SERVICE-TIME Latencies distribution in MICROSECONDS mean 10032775.17 min 9999220.74 50.00% 10066329.60 90.00% 10066329.60 99.00% 10066329.60 99.90% 10066329.60 99.99% 10066329.60 max 10066329.60 count 100000 ``` It is weird: the end-2-end latencies are stucked being ALWAYS the same with performances a little less from the PR above: probably there is something not working on message translation, please takes a look to the Open Wire tests to spot if everything is ok...
---