[SPARK-10342] Cooperative memory management

• Type: Improvement
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.5.0
• Component/s: Spark Core, SQL
• Target Version/s: 1.6.0
• Assignee: Davies Liu
• Reporter: Davies Liu
• Created: 28/Aug/15 21:03
• Updated: 30/Oct/15 21:45
• Resolved: 30/Oct/15 06:41

Description

We have memory starving problems for a long time, it become worser in 1.5 since we use larger page.

In order to increase the memory usage (reduce unnecessary spilling) also reduce the risk of OOM, we should manage the memory in a cooperative way, it means all the memory consume should be also responsive to release memory (spilling) upon others' requests.

The requests of memory could be different, hard requirement (will crash if not allocated) or soft requirement (worse performance if not allocated). Also the costs of spilling are also different. We could introduce some kind of priority to make them work together better.

Issue Links

• is duplicated by: SPARK-11348 Replace addOnCompleteCallback with addTaskCompletionListener() in UnsafeExternalSorter
• links to: [Github] Pull Request #9241 (davies)