Concurrent LRU map

From: Andrus Adamchik (andru..bjectstyle.org)
Date: Sat Dec 12 2009 - 10:36:51 EST

  • Next message: Michael Gentry: "Re: Performance Issues/Questions"

    BTW, there's no out of the box map implementation that would provide
    both concurrent non-blocking access and LRU tracking. Need to
    investigate whether we can borrow something from OSCache (IIRC it has
    a similar data structure for its main cache).

    Andrus

    On Dec 12, 2009, at 10:21 AM, Andrus Adamchik wrote:

    > On Dec 12, 2009, at 2:03 AM, halse..vv.ntnu.no wrote:
    >
    >>> http://spreadsheets.google.com/pub?key=tNytbSdt5wYIRTmTl4L-cCA&output=html
    >>
    >> If the differences are not in the database, my best guess would be
    >> that
    >> you are hitting the ugly synchronization on the DataRowStore since
    >> the
    >> problem seem to be with concurrent commits.
    >>
    >> It would be very interesting to see the same test with a quickfix on
    >> DataRowStore that removes all of the synchronization and uses stuff
    >> from
    >> java.util.concurrent instead.
    >>
    >> Regards,
    >> - Tore.
    >
    > Yeah, very likely. Also using a profiler, like YourKit should help
    > to tell why it is slowing down so much after the initial period of
    > fast operation.
    >
    > Also what's a database channel? You mean a JDBC connection?
    >
    > Andrus
    >



    This archive was generated by hypermail 2.0.0 : Sat Dec 12 2009 - 10:37:16 EST