Execute Threads + GC - Weblogic

This is a discussion on Execute Threads + GC - Weblogic ; Hi, We're noticing some degradation in performance due to GC taking place. We're trying out different tuning parameters to see what fits best but would like to know if we should be looking at other factors? For example, a colleague ...

+ Reply to Thread
Results 1 to 2 of 2

Thread: Execute Threads + GC

  1. Execute Threads + GC

    Hi,

    We're noticing some degradation in performance due to GC taking place.
    We're trying out different tuning parameters to see what fits best but
    would like to know if we should be looking at other factors?

    For example, a colleague of mine suggested that because we're running so
    many execute threads (60) this would give a longer GC than say the
    default 15. The only way I could see this happening is because WL is
    actually handling more requests simultaneously thereby creating more
    Objects that would be ready for GC. Otherwise it's all one heap that
    gets used, by all 60 threads, and only one space to clean up.

    Does anyone else have experience with this?

    We're running WLS6.1 on solaris 8

    Colin

  2. Re: Execute Threads + GC

    We have noticed similar problem during a load test on our application.
    The best approach is by finding out the GC occurances (by
    setting -verbose:gc option on the startup command line)
    and carefully increasing the YOUNG Generation Size, Survivor Ratio and
    memory size.

    What we found out is that weblogic does do create a lot of small objects per
    requests. (that is expected to be in the YOUNG Generation Area of the Heap).

    Also, Unless you are supporting more than 60 concurrent users or your
    transactions take more than a second, it is not a good idea to increase the
    number of execute threads.

    Good Luck with the tuning ....

    Thanks,
    Selva-

    "Colin Chalmers" wrote in message
    news:41b8bbe2$1@mail...
    > Hi,
    >
    > We're noticing some degradation in performance due to GC taking place.
    > We're trying out different tuning parameters to see what fits best but
    > would like to know if we should be looking at other factors?
    >
    > For example, a colleague of mine suggested that because we're running so
    > many execute threads (60) this would give a longer GC than say the default
    > 15. The only way I could see this happening is because WL is actually
    > handling more requests simultaneously thereby creating more Objects that
    > would be ready for GC. Otherwise it's all one heap that gets used, by all
    > 60 threads, and only one space to clean up.
    >
    > Does anyone else have experience with this?
    >
    > We're running WLS6.1 on solaris 8
    >
    > Colin




+ Reply to Thread