Content Management and Capture

Content Management and Capture

Come for answers. Stay for best practices. All we’re missing is you.

 View Only
  • 1.  Tuning FileNet Dispatcher for High Queue Item Volume

    Posted 7 days ago

    Hello All,

    We are experiencing a high volume of events (subscriptions) in the QueueItem table-approximately 350,000 items- per day while the server resources are performing optimally (CPU usage at 2%, RAM at 20%). The DB2 database is also functioning well.

    We would like to know how we can optimize FileNet Dispatcher to process more items from the queue efficiently.

    Specifically, could this behavior be influenced by the following parameters?

    • Under Asynchronous Processing:

      • Dispatcher wait interval

      • Maximum workers

    • Under Thread Pool Subsystem:

      • Maximum dispatcher threads

      • Maximum worker threads

    Would adjusting these settings help increase the dispatch rate for queue items?

    Appreciate your guidance on best practices or recommended values for such a high-load scenario.



    ------------------------------
    Ahmed ElHussein
    ------------------------------


  • 2.  RE: Tuning FileNet Dispatcher for High Queue Item Volume

    Posted 6 days ago
    Edited by Ronald Heerema 6 days ago

    Hi Ahmed,

    Our FileNet system is set-up to ingest up to at least 12 billion objects each year. Because of that we did a Health check of the system with IBM in order to make sure it is set-up to maximize performance.

    Based on that we changed some of the parameters you mention. In the pdf FileNet settings you can see how we configured the system.

    I also attach a guide we used to help determining our settings.

    Be aware that the tuning of the settings have a marginal effect on the total environment. It's just finetuning. Our CPE CPU's are still almost never go above 20% utilization. We do know that most performance gain can be won through optimizing the database servers where the metadata is and make sure you offload the objects into storage to keep the restrain on the database as minimal as possible as that will be the bottleneck.

    Another thing to consider is setting the workers to 1 if you have multiple CPE clusters handling your load as they will compete to update the setting of the last processed record in the database.

    I hope this will help you. Any further questions regarding the given information then just reach out.

    Regards,

    ------------------------------
    Ronald Heerema
    Architect
    Dutch Tax Office
    Apeldoorn
    +31613009685
    ------------------------------