Uploaded image for project: 'Core Server'
  1. Core Server
  2. SERVER-84002

Benchmark and compare oplog writing performance w.r.t batch size and thread number

    XMLWordPrintableJSON

Details

    • Icon: Task Task
    • Resolution: Fixed
    • Icon: Major - P3 Major - P3
    • 7.3.0-rc0
    • None
    • None
    • Fully Compatible
    • Repl 2023-12-25, Repl 2024-01-08, Repl 2024-01-22
    • 0

    Description

      This ticket should try to answer the following question by doing benchmarks for oplog writes:

      • For single thread oplog writes, how does the batch size (in terms of number of entries as well as total byte size) impact the performance?
      • For certain batch sizes (likely those big ones), what is the performance difference between writing the whole batch with 1 thread in one storage transaction vs. splitting the batch to write using multiple threads.

      The performance can also be impacted by the type of workloads (e.g. the type of the operation: insert, update, delete; the size of each operation, etc), so the benchmarks should include different types of workloads. We can even mimic some industry benchmark workloads such as the YCSB workload that are used in sys-perf.

      We should write a report to summarize the findings and benchmark results.

      Attachments

        Activity

          People

            kishore.devireddy@mongodb.com Kishore Devireddy
            wenbin.zhu@mongodb.com Wenbin Zhu
            Votes:
            0 Vote for this issue
            Watchers:
            4 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: