rich7420 commented on PR #10503: URL: https://github.com/apache/ozone/pull/10503#issuecomment-5224733096
Thanks @spacemonkd. I packaged the workflow into a small benchmark you can run and average on your side: https://github.com/rich7420/ozone/commit/a6bd20c745 Adds compose/ozone/om-write-bench/. Run from compose/ozone: N=40000 THREADS_LIST="16 64 128" ITER=3 ./om-write-bench/om-write-bench.sh Phase A sweeps single-bucket concurrency, phase B spreads the same load across N buckets (scale ratio), phase C samples OM thread CPU. freon runs in its own container; results go to results.csv / summary.txt with mean +/- stddev. For a real cluster, override FREON_RUN / OZONE_ADMIN / OM_TOP / OM_JSTACK (see README). Compare the single-vs-multi-bucket ratio, not raw keys/s. On my single-node compose (laptop, directional only): 8 buckets did not scale (~0.3-0.5x), and OMStateMachineApplyTransactionThread was the top OM CPU consumer under load. Curious what you get. -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
