1 Reply Latest reply on Jan 6, 2010 8:00 AM by Rich96

    measuring raid performance with ATTO benchmark

    Rich96

      Hello,

       

      I've been using the ATTO benchmark program to check the effect of upgrading drivers on raid 5 performance with my D975XBX2 motherboard.  This may or may not be the right place to ask a question about a non-Intel program, but I'm hoping someone has some experience to share.  I looked on several other website forums for a clue about this question, without finding anything relevant.  I found many screenshots of benchmark runs using ATTO, and they all had the 'Direct I/O' box checked.

       

      The ATTO benchmark program, ver 2.41, defaults to having the 'Direct I/O' box checked.  The help text says:

      "If this option is checked, file I/O on the test drive is performed with no system buffering or caching. Combine this option with Overlapped I/O for maximum asynchronous performance."

       

      I interpret this as meaning that direct I/O will get some of the operating system out of the way, in order to better measure just the hardware performance?

       

      Would unchecking the 'Direct I/O' box give a 'real world' measure of the performance of the hard drive?  It would seem reasonable that since applications will access the hard drive using system buffering and caching, that one should run the test in both modes, and then decide if making a tuning change helps improve performance?

       

      Any feedback will be appreciated!

       

      Thanks,  Rich.

        • 1. Re: measuring raid performance with ATTO benchmark
          Rich96

          Hello All,

          Here are some screen shots showing a comparison of Direct I/O versus non-Direct I/O options in ATTO benchmark.

           

          First, as a baseline, I checked a single 40GB IDE Pata drive connected to a D975XBX2 motherboard.  Notice that non-Direct is slower than Direct.

           

          IDE_Test.jpg

           

          Next I ran the same comparison on the same computer, this time with four hard drives setup using RAID 5.  This is with MSM 8.8 and the latest chipset drivers.

           

          Notice that the non-Direct performance is much worse than Direct, and that non-Direct write performance is less than half of the single IDE drive.

           

          RAID_Test.jpg

          Anyone else care to post some comparison results?  Particularly from D975XBX2 boards?

           

          Thanks,  Rich.