Results 1 to 8 of 8
  1. #1
    Join Date
    Mar 2004
    Posts
    7

    Unhappy Unanswered: does more disks = better performance ?

    until today i had an Oracle server (9.2 standard edition ) on a SOLARIS 8 with only 1 disk.

    the system using it performs a lot of inserts ( its a DW ) and i got about 50,000 records inserted in 40 seconds - the table also has 4 indexes
    (i have to use insert and not SQL*loader )

    now i got a new disk and added it to the machine.
    i hoped to see performace a boost but nothing.
    not when i moved the index tablespace to a seperate disk
    and not when i moved the redo log files to a seperate disk.

    i have these parameters set:
    filesystemio_options==>setall
    disk_asynch_io==>TRUE

    is there anything else i can try ? something more to configure ?

  2. #2
    Join Date
    Jan 2004
    Location
    Croatia, Europe
    Posts
    4,094
    Provided Answers: 4
    We are running Oracle 7.3 on Digital Alpha server under OpenVMS. Our database is now spread over 5 disks (which allows us to distribute data files more freely than you).
    A couple of years ago we had 3 disks and when we added another 2 I watched disk statistics for a month, and then redistributed Oracle data files in a new manner. Put all the disks under statistics again and the averages improved (I wanted all disks to have the same averages, more or less).

    I don't know which datafiles and tablespaces you have, but I think that moving index datafile to a different disk (than your user's datafile) was a good idea.
    You also mentioned redo log files; how about reallocating datafile that contains rolback segment(s) and - especially - temporary datafile?

    How many users use your server? If it is a multi-user machine, performance depends on their number and activities. Perhaps they were doing something that occupied your server and you didn't have a chance to notice improvements in performance?

  3. #3
    Join Date
    Jan 2004
    Posts
    370
    How do the performance statistics compare now you have added the extra disk?

  4. #4
    Join Date
    Mar 2004
    Posts
    7
    hi , thanks for the Reply

    the erver is dedicated to this application and the main action we perform is insert data and one or two users Query once in a while..

    i oly have 2 disks so and i thaight that moving indexes and redo files can
    perform the performace most significantlly

    i find it very strage that it did not..

    i came for the NT platform and dont have many Unix knowledge
    does anyone know how to check for CPU usage? maybe the disks are not the main problem at all...
    on an NT2000 pc i get 4 times faster response time then the Solaris8 server. i find it hard to believe...

    it feels like i m missing something...

  5. #5
    Join Date
    Mar 2004
    Posts
    7

    statistics remains exactly the same

    Originally posted by SkyWriter
    How do the performance statistics compare now you have added the extra disk?
    statistics remains exactly the same

  6. #6
    Join Date
    Mar 2002
    Location
    Reading, UK
    Posts
    1,137
    The most likely reason for it not improving is if you have another bottleneck which is actually causing the slow performance. Once you remove that bottleneck you should find the additional disk making a difference.

    Have a look at the disk stats to see if the disks are actually causing waits, if it is another bottleneck then you shouldnt see too many IO waits as it will busy doing other things probably.

    Alan

  7. #7
    Join Date
    Jan 2004
    Location
    Scottsdale, AZ
    Posts
    106
    What are the baseline measurements you used to determine if you had an improvement or not.
    "Take Control!
    Make The Choice!
    Accept The Consequences!
    Deal With The Results! "

  8. #8
    Join Date
    Oct 2003
    Posts
    706
    While having multiple disk drives is conceptually a favorable idea, there are many other factors which can work for or against it in practice.

    Multiple disks work to your advantage only when the computer is actually able to initiate meaningful I/O operations against more than one disk at the same time, or when it is able to meaningfully take advantage of the fact that the read/write head on a particular device didn't have to move (when otherwise it would have).

    In the first scenario, you are exploiting actual parallelism in the I/O subsystem. If you have something advanced, like SCSI disks, you can do that. (IDE won't.)

    In the second scenario, you're trying to avoid seek-delays, which are much slower than rotational-delays on a disk drive. (And presumably the computer's memory cache didn't eliminate the I/O operation entirely.)

    Anyhow ... a performance gain from a properly-equipped multiple-disk I/O-farm can be obtained, but it is by no means a foregone conclusion.
    ChimneySweep(R): fast, automatic
    table repair at a click of the
    mouse! http://www.sundialservices.com

Posting Permissions

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