• Icon: Task Task
    • Resolution: Done
    • Icon: Major Major
    • 1.2.0.CR1
    • None
    • test-infrastructure
    • None

      We saw repeated reports that SMTs add some non-neglectible through-put penalty. Let's expand the existing micro benchmark to get some actual numbers. Would be interesting to know:

      • What overhead adds a simple pass-through SMT?
      • What overhead adds the new filtering SMT with Groovy and GraalJS (DBZ-1782)

      Should check out what is about to be provided via DBZ-1812.

            [DBZ-1929] Add performance test for SMTs like filters

            Released

            Jiri Pechanec added a comment - Released

            Logged DBZ-2110 for implementing a true end-to-end benchmark, so to get an understanding of the SMT impact in practice.

            Gunnar Morling added a comment - Logged DBZ-2110 for implementing a true end-to-end benchmark, so to get an understanding of the SMT impact in practice.

            Mostly, yes. What we still don't have is a number for SMT overhead per se; i.e. how much latency source-to-kafka is added by putting the simplest SMT into the chain.

            Gunnar Morling added a comment - Mostly, yes. What we still don't have is a number for SMT overhead per se ; i.e. how much latency source-to-kafka is added by putting the simplest SMT into the chain.

            Jiri Pechanec added a comment - gunnar.morling Was not this done with https://github.com/debezium/debezium/pull/1366 ?

              jpechane Jiri Pechanec
              gunnar.morling Gunnar Morling
              Votes:
              0 Vote for this issue
              Watchers:
              2 Start watching this issue

                Created:
                Updated:
                Resolved: