Uploaded image for project: 'Spark'
  1. Spark
  2. SPARK-946

Reduce memory footprint of DiskBlockManager.blockToFileSegmentMap

    XMLWordPrintableJSON

Details

    • Bug
    • Status: Resolved
    • Major
    • Resolution: Fixed
    • 0.8.1, 0.9.0
    • 0.8.1, 0.9.0
    • None
    • None

    Description

      blockToFileSegmentMap right now is taking up ~400 bytes per FileSegment. In large shuffles (e.g., >1000 mappers/executor and >1000 reducers), this can lead to several GB used just for this map, which is leading to OOMs.

      Attachments

        Activity

          People

            ilikerps Aaron Davidson
            ilikerps Aaron Davidson
            Votes:
            0 Vote for this issue
            Watchers:
            1 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: