1. Get rid of all advertisements and get unlimited access to documents by upgrading to Premium Membership. Upgrade to Premium Now and also get a Premium Badge!

Has anyone verified the performance of DBMS_PARALLEL_EXECUTE package in plsql 11.2?

Discussion in 'SQL PL/SQL' started by RG Hegde, Feb 12, 2015.

  1. RG Hegde

    RG Hegde Forum Advisor

    Messages:
    185
    Likes Received:
    35
    Trophy Points:
    310
    Location:
    Pune
    Hello ,

    It is a very nice and effective concept DBMS_PARALLEL_EXECUTE from pl/sql 11.2 release .

    I have gone through the concept and executed . I succeeded in that .

    You can go through the concept http://rghegdedisc.blogspot.in/2015/01/execute-parallel-task-using.html

    I tried to verify the performance of this with some other concept like simple update, bulk update.
    But I dint get huge data to prove the performance .

    Has anyone verified this? Could you please send me the screen shot ...

    Thanks in advance
     
  2. ocprep

    ocprep Forum Advisor

    Messages:
    277
    Likes Received:
    76
    Trophy Points:
    410
    Location:
    Orlando, Florida
    Re: Has anyone verified the performance of DBMS_PARALLEL_EXECUTE package in plsql 11.

    Understand that when parallelizing jobs in Oracle, two things have to be true in order for the result to be faster than the same task.

    1. The task being executed must be large. Serial execution time should be at least several seconds long. The larger the job, the more potential benefit that executing in parallel can provide. There is overhead involved in parallel execution that can cause it to take longer than serial execution for small tasks.
    2. The server hardware must have sufficient memory and CPU resources available that parallelizing the job does not cause resource bottlenecks. As with #1, executing a job in parallel on a system with insufficient resources can cause a task to run slower than it would have serially.

    I suspect that at least #1 and possibly #2 is not true for the tests you have run. In a test environment one of the more difficult tasks is creating a sample data set that is 'large' by Oracle standards and this is pretty much required for the type of benchmark you are asking for.
     
    RG Hegde likes this.