Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Your S2PX-generated Parallel

...

Jobs may take longer to execute than the Server Jobs from which they were derived. There are a number of potential reasons for this…

  • As you’ll already be aware, the startup time for Parallel Jobs is longer than for Server Jobs, meaning low-volume jobs could behave slower. Higher volume Jobs, however, are likely to perform better due to the inherent performance benefits provided by the Parallel engine.

  • All S2PX-generated jobs use Parallel stages which are defined as running in Sequential mode , as there’s no way S2PX can infer the data necessary to generate partitioning specifications. This is easily remediated (in the generated Parallel Jobsjobs) by Developers, where there is a business case to do so.

  • For each Server job S2PX potentially generates multiple Parallel jobs for each Server job, along could potentially generate /wiki/spaces/S2PX/pages/1525579787 (each with an unavoidable startup overhead), along with a coordinating Job Sequence. Each Parallel Job introduces an unavoidable startup overhead. This decomposition is unavoidable due to stage synchronisation. (See link)

  • Hashed files beign being replaced by the DRS stage means bulk record read could be slowed, but lookups (when utilising database indices) could be faster.Anything else?