Page 2 of 2 FirstFirst 12
Results 11 to 12 of 12

Thread: A Detailed Guide To Phoronix Test Suite 2.0

  1. #11

    Default

    Quote Originally Posted by fxfuji View Post
    Is there any way to determine what the command-line equivalent is for each of the tests in the GUI implementation of PTS 2.0 ? I'd really like to know what the setting is for the record size in the IOzone test!
    It should ask you regardless of whether you are using GUI or CLI.

  2. #12
    Join Date
    Aug 2009
    Posts
    20

    Default

    Quote Originally Posted by Michael View Post
    It should ask you regardless of whether you are using GUI or CLI.
    Both ask for the size, which is the size of the file being tested. I was asking about the 'record size' parameter.

    However, your tip about the CLI did lead me to the answer. The command line invoked by PTS to iozone showed a record size of 1K.

    That value seems rather low to me... do Linux file systems typically transfer files in 1 K chunks? That seems like it would be incredibly inefficient for large files.

    I know that Windoze is altogether different, but for it, '64 K is the typical record size that Windows uses when applications try to transfer blocks of data that are bigger than 64 K' according to Don Capps of iozone:

    http://www.smallnetbuilder.com/content/view/24310/77/


    The one other iozone test I remember finding used a record size of 64 K, I believe... I'll try to dig up the link.

    *****

    Edit: Found the test I remembered, and another one.

    The one I was thinking of was from IBM, and they tested a virtual Linux machine of 192 MB with an ext2 filesystem with an 800 MB test file size. The record size they used for the test was 64 KB:

    http://www.vm.ibm.com/perf/reports/zvm/html/iozone.html


    The other one was a preliminary test of the local disk that was going to be used in a Linux cluster. The computer memory was 1 GB, the test file size was 2 GB, and a variety of filesystems were tested. They swept the record size from 64 KB to 16384 KB, in 2x steps (i.e. they doubled the record size for the subsequent test step):

    http://marc.info/?l=redhat-linux-clu...11337395629359


    It seems that the implementation of the iozone test in PTS 2.0 could benefit from being tweaked a little, using a larger record size, or even permitting the user/tester to choose the value. A larger value would probably speed up the test somewhat -- my tests using a 4 KB record size for 4 GB and 8 GB file sizes were unbearably slow, whereas using a 1 MB record size was tolerably slow. :P

    Unfortunately, I still don't know.... what is the right record size I should use for my test?
    Last edited by fxfuji; 08-15-2009 at 04:19 PM.

Posting Permissions

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