brandboat commented on PR #16305: URL: https://github.com/apache/kafka/pull/16305#issuecomment-2165051695
> look into the use case of nodeStats: the number of node should be less than number of "get" very much. Hence, we need to go back to benchmark to reduce the number of writePerLoops to test "get" again. 10% updated is a bit weird to me for this case. Test with writePercentage 0.01 (100 times of computeIfAbsent) and 0.0001 (10 times of computeIfAbsent), though there is not much difference under 0.0001, ConcurrentHashMap still performs slightly better than CopyOnWriteMap. ``` Benchmark (mapSize) (writePercentage) Mode Cnt Score Error Units ConcurrentMapBenchmark.testConcurrentHashMapGet 100 0.01 avgt 5 3.871 ± 0.149 ns/op ConcurrentMapBenchmark.testConcurrentHashMapGet 100 0.0001 avgt 5 2.259 ± 0.138 ns/op ConcurrentMapBenchmark.testConcurrentHashMapGetRandom 100 0.01 avgt 5 6.320 ± 0.231 ns/op ConcurrentMapBenchmark.testConcurrentHashMapGetRandom 100 0.0001 avgt 5 4.070 ± 0.669 ns/op ConcurrentMapBenchmark.testCopyOnWriteMapGet 100 0.01 avgt 5 57.898 ± 1.769 ns/op ConcurrentMapBenchmark.testCopyOnWriteMapGet 100 0.0001 avgt 5 2.767 ± 0.126 ns/op ConcurrentMapBenchmark.testCopyOnWriteMapGetRandom 100 0.01 avgt 5 61.361 ± 1.903 ns/op ConcurrentMapBenchmark.testCopyOnWriteMapGetRandom 100 0.0001 avgt 5 4.270 ± 0.234 ns/op ``` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: jira-unsubscr...@kafka.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
