Uploaded image for project: 'Spark'
  1. Spark
  2. SPARK-7712 Window Function Improvements
  3. SPARK-8638

Window Function Performance Improvements

    XMLWordPrintableJSON

    Details

    • Type: Sub-task
    • Status: Resolved
    • Priority: Major
    • Resolution: Fixed
    • Affects Version/s: None
    • Fix Version/s: 1.5.0
    • Component/s: SQL
    • Labels:
      None
    • Target Version/s:

      Description

      Improve the performance of Spark Window Functions in the following cases:

      1. Much better performance (10x) in the running case (e.g. BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW). The current implementation in spark uses a sliding window approach in these cases. This means that an aggregate is maintained for every row, so space usage is N (N being the number of rows). This also means that all these aggregates all need to be updated separately, this takes N*(N-1)/2 updates. The running case differs from the Sliding case because we are only adding data to an aggregate function (no reset is required), we only need to maintain one aggregate (like in the UNBOUNDED PRECEDING AND UNBOUNDED case), update the aggregate for each row, and get the aggregate value after each update. This is what the new implementation does. This approach only uses 1 buffer, and only requires N updates; I am currently working on data with window sizes of 500-1000 doing running sums and this saves a lot of time.
        #. Fewer comparisons in the sliding case. The current implementation determines frame boundaries for every input row. The new implementation makes more use of the fact that the window is sorted, maintains the boundaries, and only moves them when the current row order changes. This is a minor improvement.
      2. A single Window node is able to process all types of Frames for the same Partitioning/Ordering. This saves a little time/memory spent buffering and managing partitions. This will be enabled in a follow-up PR.
      3. A lot of the staging code is moved from the execution phase to the initialization phase. Minor performance improvement, and improves readability of the execution code.

      The attached perf_test.scala file contains s number of queries which can be used to measure the differences between the current and the proposed window function implementation. In the tests the new implementation outperforms the current implementation by a factor 7x in sliding window cases, and by a factor 14x in the running window cases.

        Attachments

        1. perf_test.scala
          3 kB
          Herman van Hövell
        2. perf_test2.scala
          4 kB
          Herman van Hövell
        3. perf_test3.scala
          6 kB
          Herman van Hövell

          Activity

            People

            • Assignee:
              hvanhovell Herman van Hövell
              Reporter:
              hvanhovell Herman van Hövell
              Shepherd:
              Yin Huai
            • Votes:
              0 Vote for this issue
              Watchers:
              6 Start watching this issue

              Dates

              • Created:
                Updated:
                Resolved: