Results 1 to 2 of 2

Thread: RMS performance improvement

  1. #1
    Tzachi Nissim is offline Product Implementation Director
    Join Date
    Aug 2006
    Rep Power

    RMS performance improvement

    As part of OpenVMS 7.2 1H1 a new feature was added to RMS that can improve data access performance to RMS. This feature deals with eliminating calls to the VMS lock manager in specific situations. AIS turns on the NLK and RRL RMS flags (“no locking” and “read regardless of locks”) when reading from RMS files. Despite that, by default RMS still calls the RMS lock manager to establish the lock status of a record. This is very costly and completely unnecessary in a read scenario. The new VMS feature allows this overhead to be eliminated.

    The following command needs to be incorporated in the server’s environment (normally done in NAVROOT:[BIN]SITE_NAV_LOGIN.COM):


    Our tests and customer testimonies have shown roughly 30% performance improvement by taking advantage of this feature.
    To Find Out more About Attunity Technology:
    Contact Us

  2. #2
    Hein is offline Senior Member
    Join Date
    Dec 2007
    Nashua, NH - USA.
    Rep Power

    RMS RAB$V_NQL = No-Query Lock

    Right, RMS will still take out a record lock even if RNL and/or RRL is specified. This is required to satisfy certain language (COBOL!) requirements, but it is generaly unexpected and undesirable.

    RMS 'properly' addressed this with a new option NQL (No Query Lock)
    The Attunity RMS driver should really be using that option. Period.

    will 'convert' any RNL+RRL to be interpreted as NQL, but that's almost just a workaround for an actively developed/supported product.

    I'd recommend using this new switch once and for all using the same command with /SYSTEM added. No need to update the NAV_LOGIN in that case.

    Note that for every record RMS still will, at the very minimum, do the convert up and down for the bucket lock. That's for simple sequential primary key access. For indexed access it is more likely it will do 8+ converts as it drills down that index tree (this can be mitigated with Global Buffers). Thus the beneficial effect will not likely be close to 30% for keyed access nor for alternate key sequential read access.

    Hein van den Heuvel
    HvdH Performance Consulting.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts