[FLINK-28828] Sorting all unfinished readers in batches at one time in SortMergeResultPartitionReadScheduler

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.16.0
• Component/s: Runtime / Network
• Labels: pull-request-available
• Assignee: Yuxin Tan
• Reporter: Yuxin Tan
• Created: 05/Aug/22 07:22
• Updated: 08/Aug/22 07:18
• Resolved: 07/Aug/22 12:53

Description

Currently, when reading data in SortMergeResultPartitionReadScheduler, the reader is added to the priority queue immediately. However, the data read from this reader may not have been consumed, which will cause this reader to be ranked later in the queue, which is unfavorable to sequential reading.

To solve the issue, After reading the data, we should sort all unfinished readers in batches at one time, that is, add all unfinished readers to the priority queue, which is more conducive to sequential reading.

Issue Links

• links to: GitHub Pull Request #20463