Results 1 to 4 of 4

Thread: I/O Wait time

  1. #1
    Join Date
    Mar 2005
    Posts
    73

    Unanswered: I/O Wait time

    UDB 8.2 FP7 OS: AIX.

    Friends.

    I am seeing lot of I/o wait times on my system. Any my users complaining of slow performance. I am not sure how to solve this problem.
    Can you some one give me tips on how to start diggiing the issue.

    THe mpstat command shows lot of wait times and the TOP on averate give 50 percent I/O wait.

    CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys wt idl
    0 0 0 23 4 1 21 0 1 0 0 27 0 0 0 100
    1 0 0 46 804 1 29 0 0 0 0 32 0 0 0 100
    2 0 0 1 4 1 24 0 0 0 0 2 0 0 0 100
    3 0 0 1 956 1 26 0 0 0 0 19 0 1 0 99
    4 0 0 25 95 92 234 0 0 0 0 124 1 1 92 6
    5 0 0 9 77 74 104 0 0 0 0 100 1 1 6 92
    6 0 0 1 101 99 219 0 0 0 0 123 1 1 98 0
    7 0 0 46 348 246 30 0 0 0 0 26 0 0 0 100


    load averages: 0.33, 0.40, 0.49; up 104+15:56:42 14:39:57
    493 processes: 492 sleeping, 1 on cpu
    CPU states: 46.0% idle, 2.2% user, 2.3% kernel, 49.4% iowait, 0.0% swap
    Memory: 16G phys mem, 8432M free mem, 16G swap, 16G free swap

  2. #2
    Join Date
    Sep 2003
    Posts
    237
    One needs to look at topas or nmon output. AIX parameters minperm,maxperm,maxpgahead(jfs2 vesion if you are using jfs2)need to be tweaked to match prefetch parameters of tablespaces; also the number of prefetchers, numiocleaners etc. has to be adjusted. Also find out which SQLs are causing slowdown; make sure they are not using full table scans(you can do this by snapshot monitoring). See if appropriate indexes can be created to reduce full table scans. May be create seperate bufferpools for index and data tablespaces; also allocate blocks in each tablespace to accomodate big block fetches.
    mota

  3. #3
    Join Date
    Mar 2005
    Posts
    73
    Thanks dbamota.

    This is a solaris environment. My mpstat or vmstat doesn't show any high cpu utilization. But i see a lot of i/o wait for cpu s in the top command.
    I/o waits are on average 50 percent and goes to as high as 85-90 percent.

    What are the other things i can look at ? SQL tuning?

    Can any one guide me how to proceed from here.

    thanks

  4. #4
    Join Date
    Sep 2003
    Location
    canada
    Posts
    230
    Quote Originally Posted by dba_udb
    Thanks dbamota.

    This is a solaris environment. My mpstat or vmstat doesn't show any high cpu utilization. But i see a lot of i/o wait for cpu s in the top command.
    I/o waits are on average 50 percent and goes to as high as 85-90 percent.

    What are the other things i can look at ? SQL tuning?

    Can any one guide me how to proceed from here.

    thanks
    1-in application side do reorge, runstats and bind database
    2- look at the sql statements and create proper indexes.

Posting Permissions

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