DB2 8.1.4 I/O Performance

Collapse
This topic is closed.
X
X
 
  • Time
  • Show
Clear All
new posts
  • joerg

    DB2 8.1.4 I/O Performance

    Hello,
    we have quite a lot of troubles with our new installation of DB2 8.1.4
    on a Linux-System. The server is a Dual-Xeon with 2.4 GHz and 2GB Ram,
    with installed Raid 1 mirrored 75 GB harddisks.
    The system itself is performing very fast but in combination with DB2
    we are facing an I/O-problem.
    Our DB-scenaio is as follows:

    Loading a flat table over JDBC and splitting it into a more detailed
    datastructure. We are commiting after every successful transformation
    (no autocommit). What we now see is that the DB is doing almost
    nothing and the server itself is also more or less idle.

    We check everything we thought of and we found the following things:

    up to 4 blocked processes (see VMSTAT result)
    very low buffer-out rate (see VMSTAT result)
    terribly high time of asynchronous write (see BUFFERPOOL
    snapshot)

    Has anbody an idea how to solve this problem ?

    Thanks

    Joerg



    Result of VMSTAT

    procs -----------memory---------- ---swap-- -----io---- --system--
    ----cpu----
    r b swpd free buff cache si so bi bo in cs us
    sy id wa
    0 2 29156 75196 217096 2063308 0 0 17 33 17 9 1
    1 98 0
    0 1 29156 75196 217104 2063656 0 0 0 930 803 945 3
    3 94 0
    0 3 29156 73700 217104 2063816 0 0 0 895 770 910 3
    1 96 0
    0 4 29156 73700 217104 2064144 0 0 2 926 783 946 2
    1 97 0

    Result of BUFFERPOOL-snapshot

    Buffer pool data logical reads = 1368786
    Buffer pool data physical reads = 40172
    Buffer pool temporary data logical reads = 0
    Buffer pool temporary data physical reads = 0
    Buffer pool data writes = 18557
    Buffer pool index logical reads = 21874494
    Buffer pool index physical reads = 686
    Buffer pool temporary index logical reads = 0
    Buffer pool temporary index physical reads = 0
    Total buffer pool read time (ms) = 13395
    Total buffer pool write time (ms) = 1792986
    Asynchronous pool data page reads = 27476
    Asynchronous pool data page writes = 15658
    Buffer pool index writes = 19106
    Asynchronous pool index page reads = 0
    Asynchronous pool index page writes = 11751
    Total elapsed asynchronous read time = 3329
    Total elapsed asynchronous write time = 1308907
    Asynchronous data read requests = 1718
    Asynchronous index read requests = 0
    No victim buffers available = 18560
    Direct reads = 0
    Direct writes = 0
    Direct read requests = 0
    Direct write requests = 0
    Direct reads elapsed time (ms) = 0
    Direct write elapsed time (ms) = 0
    Database files closed = 3
    Data pages copied to extended storage = 0
    Index pages copied to extended storage = 0
    Data pages copied from extended storage = 0
    Index pages copied from extended storage = 0
    Unread prefetch pages = 0
    Vectored IOs = 1718
    Pages from vectored IOs = 27476
    Block IOs = 0
    Pages from block IOs = 0
    Physical page maps = 0
  • Mark Yudkin

    #2
    Re: DB2 8.1.4 I/O Performance

    Sounds like your disk controller. Try adding battery-backed memory to the
    SCSI RAID controller, or updating the RAID controller card to one that can
    manage better write performance.

    (I've assume you have SCSI RAID as you said Dual Xeon and RAID-1. If you
    don't, that's probably your problem).

    BTW, how large are your buffer pools? (If they're too small that can also
    produce problems, but probably isn't your problem here).

    "joerg" <joerg.sailer@p ublica.de> wrote in message
    news:475a2480.0 312042345.4135b 967@posting.goo gle.com...[color=blue]
    > Hello,
    > we have quite a lot of troubles with our new installation of DB2 8.1.4
    > on a Linux-System. The server is a Dual-Xeon with 2.4 GHz and 2GB Ram,
    > with installed Raid 1 mirrored 75 GB harddisks.
    > The system itself is performing very fast but in combination with DB2
    > we are facing an I/O-problem.
    > Our DB-scenaio is as follows:
    >
    > Loading a flat table over JDBC and splitting it into a more detailed
    > datastructure. We are commiting after every successful transformation
    > (no autocommit). What we now see is that the DB is doing almost
    > nothing and the server itself is also more or less idle.
    >
    > We check everything we thought of and we found the following things:
    >
    > up to 4 blocked processes (see VMSTAT result)
    > very low buffer-out rate (see VMSTAT result)
    > terribly high time of asynchronous write (see BUFFERPOOL
    > snapshot)
    >
    > Has anbody an idea how to solve this problem ?
    >
    > Thanks
    >
    > Joerg
    >
    >
    >
    > Result of VMSTAT
    >
    > procs -----------memory---------- ---swap-- -----io---- --system--
    > ----cpu----
    > r b swpd free buff cache si so bi bo in cs us
    > sy id wa
    > 0 2 29156 75196 217096 2063308 0 0 17 33 17 9 1
    > 1 98 0
    > 0 1 29156 75196 217104 2063656 0 0 0 930 803 945 3
    > 3 94 0
    > 0 3 29156 73700 217104 2063816 0 0 0 895 770 910 3
    > 1 96 0
    > 0 4 29156 73700 217104 2064144 0 0 2 926 783 946 2
    > 1 97 0
    >
    > Result of BUFFERPOOL-snapshot
    >
    > Buffer pool data logical reads = 1368786
    > Buffer pool data physical reads = 40172
    > Buffer pool temporary data logical reads = 0
    > Buffer pool temporary data physical reads = 0
    > Buffer pool data writes = 18557
    > Buffer pool index logical reads = 21874494
    > Buffer pool index physical reads = 686
    > Buffer pool temporary index logical reads = 0
    > Buffer pool temporary index physical reads = 0
    > Total buffer pool read time (ms) = 13395
    > Total buffer pool write time (ms) = 1792986
    > Asynchronous pool data page reads = 27476
    > Asynchronous pool data page writes = 15658
    > Buffer pool index writes = 19106
    > Asynchronous pool index page reads = 0
    > Asynchronous pool index page writes = 11751
    > Total elapsed asynchronous read time = 3329
    > Total elapsed asynchronous write time = 1308907
    > Asynchronous data read requests = 1718
    > Asynchronous index read requests = 0
    > No victim buffers available = 18560
    > Direct reads = 0
    > Direct writes = 0
    > Direct read requests = 0
    > Direct write requests = 0
    > Direct reads elapsed time (ms) = 0
    > Direct write elapsed time (ms) = 0
    > Database files closed = 3
    > Data pages copied to extended storage = 0
    > Index pages copied to extended storage = 0
    > Data pages copied from extended storage = 0
    > Index pages copied from extended storage = 0
    > Unread prefetch pages = 0
    > Vectored IOs = 1718
    > Pages from vectored IOs = 27476
    > Block IOs = 0
    > Pages from block IOs = 0
    > Physical page maps = 0[/color]


    Comment

    • PM \(pm3iinc-nospam\)

      #3
      Re: DB2 8.1.4 I/O Performance

      NUM_IOCLEANERS = ?

      Command :
      get db cfg for dbname

      You can also get a few snapshots to investigate
      *log*
      Dirty page steal cleaner triggers
      Agents stolen
      Database files closed
      Number of files closed
      etc.


      PM

      "joerg" <joerg.sailer@p ublica.de> a écrit dans le message de
      news:475a2480.0 312042345.4135b 967@posting.goo gle.com...[color=blue]
      > Hello,
      > we have quite a lot of troubles with our new installation of DB2 8.1.4
      > on a Linux-System. The server is a Dual-Xeon with 2.4 GHz and 2GB Ram,
      > with installed Raid 1 mirrored 75 GB harddisks.
      > The system itself is performing very fast but in combination with DB2
      > we are facing an I/O-problem.
      > Our DB-scenaio is as follows:
      >
      > Loading a flat table over JDBC and splitting it into a more detailed
      > datastructure. We are commiting after every successful transformation
      > (no autocommit). What we now see is that the DB is doing almost
      > nothing and the server itself is also more or less idle.
      >
      > We check everything we thought of and we found the following things:
      >
      > up to 4 blocked processes (see VMSTAT result)
      > very low buffer-out rate (see VMSTAT result)
      > terribly high time of asynchronous write (see BUFFERPOOL
      > snapshot)
      >
      > Has anbody an idea how to solve this problem ?
      >
      > Thanks
      >
      > Joerg
      >
      >
      >
      > Result of VMSTAT
      >
      > procs -----------memory---------- ---swap-- -----io---- --system--
      > ----cpu----
      > r b swpd free buff cache si so bi bo in cs us
      > sy id wa
      > 0 2 29156 75196 217096 2063308 0 0 17 33 17 9 1
      > 1 98 0
      > 0 1 29156 75196 217104 2063656 0 0 0 930 803 945 3
      > 3 94 0
      > 0 3 29156 73700 217104 2063816 0 0 0 895 770 910 3
      > 1 96 0
      > 0 4 29156 73700 217104 2064144 0 0 2 926 783 946 2
      > 1 97 0
      >
      > Result of BUFFERPOOL-snapshot
      >
      > Buffer pool data logical reads = 1368786
      > Buffer pool data physical reads = 40172
      > Buffer pool temporary data logical reads = 0
      > Buffer pool temporary data physical reads = 0
      > Buffer pool data writes = 18557
      > Buffer pool index logical reads = 21874494
      > Buffer pool index physical reads = 686
      > Buffer pool temporary index logical reads = 0
      > Buffer pool temporary index physical reads = 0
      > Total buffer pool read time (ms) = 13395
      > Total buffer pool write time (ms) = 1792986
      > Asynchronous pool data page reads = 27476
      > Asynchronous pool data page writes = 15658
      > Buffer pool index writes = 19106
      > Asynchronous pool index page reads = 0
      > Asynchronous pool index page writes = 11751
      > Total elapsed asynchronous read time = 3329
      > Total elapsed asynchronous write time = 1308907
      > Asynchronous data read requests = 1718
      > Asynchronous index read requests = 0
      > No victim buffers available = 18560
      > Direct reads = 0
      > Direct writes = 0
      > Direct read requests = 0
      > Direct write requests = 0
      > Direct reads elapsed time (ms) = 0
      > Direct write elapsed time (ms) = 0
      > Database files closed = 3
      > Data pages copied to extended storage = 0
      > Index pages copied to extended storage = 0
      > Data pages copied from extended storage = 0
      > Index pages copied from extended storage = 0
      > Unread prefetch pages = 0
      > Vectored IOs = 1718
      > Pages from vectored IOs = 27476
      > Block IOs = 0
      > Pages from block IOs = 0
      > Physical page maps = 0[/color]


      Comment

      Working...