DBAsupport.com Forums - Powered by vBulletin
Results 1 to 6 of 6

Thread: log file parallel write

  1. #1
    Join Date
    Jun 2001
    Posts
    193
    what should i do if i found event

    log file sequential read
    log file parallel write
    log buffer space

    a lot
    guru is on the way!!!!

  2. #2
    Join Date
    Nov 2000
    Location
    greenwich.ct.us
    Posts
    9,092
    Check http://otn.oracle.com/docs/products/...pa3.htm#264253 for the list of wait events.

    To look at what each event means, look at http://otn.oracle.com/docs/products/...76961/apa5.htm

    Excessive counts and times for these indicate you probably have contention in your logfiles.
    Jeff Hunter

  3. #3
    Join Date
    Jun 2001
    Posts
    193
    marist89:

    just like what i said before in another thread.

    i am the only user of the system.
    i bring up only ONE sqlplus window.
    and run tom's example:

    begin
    for i in 1..100000
    loop
    insert into t values(i);
    commit;
    end loop;
    end;




    i find out there are log file parallel write event.

    can u tell me how to solve it?
    i know it is because LGWR can't keep up, but don't know how to solve it.


    does that mean my system is i/o boundary, and your system is not, since you can run 5 tom's example at the same time without log file parallel write event, mine have log file parallel write event even only one process is running.

    does that explain why I get some for data block, free list, segment header, undo block, undo header, etc.


    guru is on the way!!!!

  4. #4
    Join Date
    Apr 2001
    Posts
    219
    Your LGWR has to write out evey commit. So you can reduce the frequency by commiting every nth record and possibly increasing I/O performance. But to tell if you are really effectted by I/O look at the disks that hold the redo logs. On Solaris I use "iostat -xnc 3", along with "top" and "vmstat 3". This helps me figure out my I/O problems quick.

  5. #5
    Join Date
    Jun 2001
    Posts
    193
    so there is no magic init.ora parameter can speed lgwr?
    guru is on the way!!!!

  6. #6
    Join Date
    Apr 2001
    Posts
    219
    Well there are parameters that effect checkpoints and log buffer size which can effect the LGWR. But, you are doing large amount of inserts in a small amount of time, which forces commits to disk immediately. You can increase performance by doing your inserts in parallel, that is if you have multiple CPUs and enough redo log I/O. But if your machine does not have enough resource to handle parallel execution then you are stuck.

    I have my redo logs on multiple RAID 0 mount points, becuase my largest DB handle around a million inserts an hour. So, I write any where from 1.5MB to 5MB a second to disk and my disks have low latency and busy usage. My smallest-heavy machine is restricted by CPU, not I/O, so if we get more CPUs my system will scale to handle more load. I also had the programmers write there program to do parallel inserts, which works great.

Posting Permissions

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


Click Here to Expand Forum to Full Width