poor performance due to bad disc file distribution

Hi.
I'm in trouble with ver. 3.56a.
My Unix sistem runs navision and the database file is distributed across 6 logical disc. Well, one month ago the write average time was fine with more or less the same time amount in all of disc (42 ms) and the performance was very good. Today we have the first disc in 120 ms and the other remains in 42 ms. This one produces a poor performnace of the system.

Is there anyone that could know anything about it.
Thanks a lot.



  • I good mean write time isd just about the disks access time, That means between 5 and 12 ms.

    your system has severel problems.

    1. the logical disks doesn't give you any perfomance at all. You should distribute the database, betweeen a number of physical disks.

    2. Something is very wrong with 42 ms. Either your disks is VERY old or the disksystem is configured wrong. ALWAYS USE RAID 1. NEVER USE RAID 0 or 5.

    The reason for the sudden change can be, that a lot of data in part one has been deleted or moved, and now the DBMS system is trying to distribute the data equally (in this case a lot of data in part 1)

    Solution

    Make a complete backup
    Test that the backup can be restored
    Reconfigure you disksystem with RAID1
    Create a new database with as many parts as physical disks
    Restore the backup.

    Now data will be randomly distributed in all parts an performance will increase significantly.

  • Hi MIchel,
    Thanks a lot for the information.

    Of course my system has great problems...
    Talking about point one we have a RAID 1 system that uses 6 physical disc those physical disc are handling as 6 logical disc.

    The only one thing that you could consider is that the one month ago we purchased a database space (500 MB). This space was applied to the database in some specific quantity just in order to equilibrate the amount of space given to all of the disc.
    We performed this task in 6 days beggining with the disc one.

    In the another hand, our database has 8.5 GB and I think that the process of performing a database backup and the restore it could be very long in terms of time (several days).

    Do you think that this activitie could be cause of this problem?

    Do you know another technique to enforce the data will be placed equally in the disk without database backup-restore?

    Thanks for the cooperation.
    Eduardo

Related