Uploaded image for project: 'Apache Tez'
  1. Apache Tez
  2. TEZ-4067

Tez Speculation decision is calculated on each update by the dispatcher

    XMLWordPrintableJSON

Details

    • Improvement
    • Status: Resolved
    • Minor
    • Resolution: Fixed
    • None
    • 0.10.0, 0.9.3
    • None
    • None

    Description

      LegacySpeculator is an object field in VertexImpl. Therefore, all events are handled synchronously by the caller (dispatcher). This implies the following:

      1. the dispatcher spends long time executing updateStatus as it needs to check the runtime estimation of the tezAttempts within the vertex.
      2. the speculator is per stage: lunching a speculation may not the optimum decision. Ideally, based on resources, speculated tasks should be the ones with slowest progress.
      3. the time between speculation is skewed because there is a big delay for the dispatcher to complete a full cycle. Also, speculation will be more aggressive compared to MR because MR waits for "soonest.retry.after.speculate" whenever a task is speculated. On the other hand, Tez speculates more tasks as it processes stages in parallel.

       

      Attachments

        1. TEZ-4067.001.patch
          23 kB
          Ahmed Hussein
        2. TEZ-4067.002.patch
          26 kB
          Ahmed Hussein
        3. TEZ-4067.003.patch
          26 kB
          Ahmed Hussein
        4. TEZ-4067.004.patch
          26 kB
          Ahmed Hussein
        5. TEZ-4067.005.patch
          28 kB
          Ahmed Hussein
        6. TEZ-4067.006.patch
          34 kB
          Ahmed Hussein
        7. TEZ-4067.007.patch
          36 kB
          Ahmed Hussein
        8. TEZ-4067.008.patch
          36 kB
          Ahmed Hussein

        Issue Links

          Activity

            People

              ahussein Ahmed Hussein
              ahussein Ahmed Hussein
              Votes:
              0 Vote for this issue
              Watchers:
              2 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: