Leave feedback
  • Question

    Performance Enhancement in Page Out/Story Teller

Enter a new topic
  • Nikhil Bansal Nikhil Bansal
    0 likes 902 views

    Hi,

    We are using SP5 build 1450. 
    We need to enhance statement generation process in our application. We have a StreamIN event(we have single input file with data of multiple users) & page out process for PDF generation(which soon we are going to upgrade to storyteller).

    Currently our project is running without queues. We did a small test in our application by enabling Queues in our project(Screenshot attached).
    Following is the result of the test : 

    APPLICATION NAME TOTAL STATEMENT COUNT TOTAL TIME TAKEN Existing Without Queues 5987 0:58:22 With Queues 5987 0:58:42

    So as per the statistics there is no change in the time taken in statement generation.

    Please suggest how can we enhance performance to generate statement?

    Regards,
    Nikhil

    Monday 02 June, 2014

    Attached files

  • Andreas Hjelle Andreas Hjelle
    2 likes

    Hi,

    Queues could enhance performance. But they might also slow down things. You will need to connect to the repository a lot and I/O capabilities on the DB will be limiting. Blob on disk should in most cases be the fastest queues, but is not available in all scenarios.

    I notice that you also have added multiple threads on the input and output queues. In your scenario, with one single input file. This will not have any effect at all. Each thread can only process one job.

    The first thing would be to analyze the job using the profiler. This will give you an idea of where the bottlenecks are. It could be internal in the streamserver (project layout, scripting and so on), but it could also be disks, EAI shares, print spoolers, network, SMTP servers and other peripherals.

    The next thing to do to increase performance would be to upgrade to 5.6.1. This is in my experience by far the fastest 5.x version. Here you might be able to utilize the job scaling functionality, a form of “job threading”, based on how your input is organized.

     

    Regards,

    andreas

    Monday 02 June, 2014