Uploaded image for project: 'Debezium'
  1. Debezium
  2. DBZ-9030

LogMiner performance regression with buffered implementation

XMLWordPrintable

      After the rework in DBZ-8860, it seems there is a small performance regression with adding new transactions to the cache:

      Benchmark                                          Mode  Cnt   Score   Error   Units
      MemoryTransactionCachePerfTest.addEventNew        thrpt   10  22.286 ± 1.950  ops/us
      MemoryTransactionCachePerfTest.addEventOld        thrpt   10  14.831 ± 2.588  ops/us
      MemoryTransactionCachePerfTest.addTransactionNew  thrpt   10   4.020 ± 0.625  ops/us
      MemoryTransactionCachePerfTest.addTransactionOld  thrpt   10   8.417 ± 2.096  ops/us
      

      With the new changes I proposed in the attached PR, performance is now:

      Benchmark                                          Mode  Cnt   Score   Error   Units
      MemoryTransactionCachePerfTest.addEventNew        thrpt   10  21.440 ± 2.633  ops/us
      MemoryTransactionCachePerfTest.addEventOld        thrpt   10  14.707 ± 2.544  ops/us
      MemoryTransactionCachePerfTest.addTransactionNew  thrpt   10   8.388 ± 1.420  ops/us
      MemoryTransactionCachePerfTest.addTransactionOld  thrpt   10   8.393 ± 1.863  ops/us
      

              ccranfor@redhat.com Chris Cranford
              ccranfor@redhat.com Chris Cranford
              Votes:
              0 Vote for this issue
              Watchers:
              2 Start watching this issue

                Created:
                Updated:
                Resolved: