1. You are viewing our forum as a guest. For full access please Register. WindowsBBS.com is completely free, paid for by advertisers and donations.

Current disk que spikes when copying large file

Discussion in 'Windows Server System' started by philinperth, 2010/04/21.

  1. 2010/04/21
    philinperth

    philinperth Inactive Thread Starter

    Joined:
    2010/04/21
    Messages:
    4
    Likes Received:
    0
    I have been monitoring a windows 2003 std server R2 (file server) using perfmon for a few weeks now....I noticed something odd when I copied a 589MB file to the file server disk array (raid 5 3 dsks SATA 2 7200rpm on Adptec Ultra 320 SCSI RAID controller riser card).....I am monitoring the Physical disk array with the counter "current disk que length" (scale 1). When the copying starts (using windows explorer) this counter spikes well in excess of 35.....

    Googling tells me this counter should be below 2 per disk so below 6 in my case for 3 dsk raid 5. I don't think I should be seeing this....but I don't understand whats happening or why this is occuring......Can anyone explain this? Network is 1Gb/s client/server TCP/IP MS client network. Wired link goes through 48port 1Gb cisco managed switch (2960G)to server which has 4 network cards teamed together.
     
  2. 2010/04/21
    amdace

    amdace Inactive

    Joined:
    2009/09/08
    Messages:
    70
    Likes Received:
    3
    Sounds like normal operation to me. You should see the counters go up when copying data. The current disk queue length is the number of system requests waiting for disk access. Since you are copying data at a sustained rate, the data is buffered and the counter goes up.
    Here is a Technet article that gives a good explanation of Disk Usage Monitoring. It should help you find out whether or not you are experiencing performance bottlenecks with regard to your disk I/O.
     

  3. to hide this advert.

  4. 2010/05/13
    MadhurjyaBora

    MadhurjyaBora Inactive

    Joined:
    2010/05/13
    Messages:
    8
    Likes Received:
    0
    You don't need to worry if these are just spikes, you will be able to better comprehend the state of the system if you create a perfmon counter log and monitor it over a period of time, for disk monitoring make sure that the sampel interval is not more than 2 secs. secondly, the most important counters to monitor the disk subsystem is Average Disk sec/transfer (this one is the most important), and the second one is avg. disk queue length. ideally avg. disk queue length is good if its around 2 per spindle (hard disk), its a concern only if yiu see this counter high for a sustained period of time. For avg. disk sec/transfer, its rule of thumb -- (Excelent: <=.012, Good <=.020, fair: <=.030, Poor: <=.040 or more)

    Hope this answers your question. :)

    - Madhurjya
     

Share This Page

  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.