[ https://issues.apache.org/jira/browse/FLINK-32306?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17731351#comment-17731351 ]
Zhu Zhu edited comment on FLINK-32306 at 6/11/23 2:03 PM: ---------------------------------------------------------- Thanks for reporting this! [~martijnvisser] This happens because we have changed these benchmarks to benchmark AdaptiveBatchScheduler(previously DefaultScheduler), which is currently the recommended and default scheduler for batch jobs. See FLINK-30480. So this is not a blocker issue. It's just some existing problems are newly exposes. Here are some more details: 1. The regression of InitScheduling.BATCH is small (several milli-seconds) and it's a one time operation, and therefore can be ignored. 2. We are aware of the performance regression of schedulingDownstreamTasks.BATCH, FLINK-32288 was opened for it and there is a fix in development. 3. The regression of startScheduling.BATCH was not expected. We will take a look. Yet it may be due to some implementation differences between the two schedulers and is acceptable. (the regression is less than 100ms, while it's a one-time operation in job lifecycle). cc [~xiasun] was (Author: zhuzh): Thanks for reporting this! [~martijnvisser] This happens because we have changed these benchmarks to benchmark AdaptiveBatchScheduler(previously DefaultScheduler), which is currently the recommended and default scheduler for batch jobs. So this is not a blocker issue. It's just some existing problems are newly exposes. Here are some more details: 1. The regression of InitScheduling.BATCH is small (several milli-seconds) and it's a one time operation, and therefore can be ignored. 2. We are aware of the performance regression of schedulingDownstreamTasks.BATCH, FLINK-32288 was opened for it and there is a fix in development. 3. The regression of startScheduling.BATCH was not expected. We will take a look. Yet it may be due to some implementation differences between the two schedulers and is acceptable. (the regression is less than 100ms, while it's a one-time operation in job lifecycle). cc [~xiasun] > Multiple batch scheduler performance regressions > ------------------------------------------------ > > Key: FLINK-32306 > URL: https://issues.apache.org/jira/browse/FLINK-32306 > Project: Flink > Issue Type: Bug > Components: Runtime / Coordination > Reporter: Martijn Visser > Priority: Blocker > > InitScheduling.BATCH > http://codespeed.dak8s.net:8000/timeline/#/?exe=5&ben=initSchedulingStrategy.BATCH&extr=on&quarts=on&equid=off&env=2&revs=200 > schedulingDownstreamTasks.BATCH > http://codespeed.dak8s.net:8000/timeline/#/?exe=5&ben=schedulingDownstreamTasks.BATCH&extr=on&quarts=on&equid=off&env=2&revs=200 > startScheduling.BATCH > http://codespeed.dak8s.net:8000/timeline/#/?exe=5&ben=startScheduling.BATCH&extr=on&quarts=on&equid=off&env=2&revs=200 -- This message was sent by Atlassian Jira (v8.20.10#820010)