Uploaded image for project: 'IMPALA'
  1. IMPALA
  2. IMPALA-9042 Support reading full-ACID ORC tables
  3. IMPALA-10903

Do a PARTITIONED ANTI HASH JOIN when there are lots of deleted rows

    XMLWordPrintableJSON

Details

    • Sub-task
    • Status: Open
    • Major
    • Resolution: Unresolved
    • None
    • None
    • Frontend
    • ghx-label-1

    Description

      Impala currently always does a BROADCAST join between base/delta files and delete delta files:

      https://github.com/apache/impala/blob/45d3eddc056bc28ee43efa6307cd673062936a41/fe/src/main/java/org/apache/impala/planner/SingleNodePlanner.java#L1726

      When there are lots of deleted rows a PARTITIONED hash join can be more efficient.

      Attachments

        Activity

          People

            Unassigned Unassigned
            boroknagyz Zoltán Borók-Nagy
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

            Dates

              Created:
              Updated: