risdenk commented on PR #1184: URL: https://github.com/apache/solr/pull/1184#issuecomment-1320600677
So I reran this update especially after f1fe35c but included the review changes as well. Memory allocation is no longer dominated by getProcessedFilter. I'd share screenshots, but custom scorers now dominate the flamegraph after cleaning this up. Some performance numbers from a cluster test: * median latency baseline from prod - something like 23ms. iteration prior to f1fe35c -> 18ms. second iteration after f1fe35c and cleanup -> 11ms * p75 in the 16ms range vs ~60ms range from prod * p99 in the 100ms range vs ~325ms range from prod * prod was hitting ~75% cpu. this is 30% CPU * GC counts were in the 250 range for prod at ~2.25s. during this test maxing around 75 range and around 500ms max * total estimated memory allocation in prod for 5 min period - 1.5TB. total memory allocated in test during 5 min period = 356GB * ~50% reduction in latency in ms across the board for percentiles under 99.99%. -- 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]
