No announcement yet.

Striped drives produce low numbers

  • Filter
  • Time
  • Show
Clear All
new posts

  • Striped drives produce low numbers

    I am testing some Seagate SATA drives in a system. I have a single drive as a system drive C: and two more that are 'striped' using the MS disk management as a single drive D:.

    I am running the advanced disk test, set for 2048MB file size and 2048 byte block size, Win32API - Cached, 90% read and 90% sequential.

    The single drive is reporting anywhere from 5 to 10 times faster data transfer than the striped pair of drives. This does not make a lot of sense, since I know that the striped drives are generally faster.

    Is there some magic required for testing a striped set of drives?

  • #2
    Larger block sizes should give better performance.

    It would be interesting to know if the drives in the software RAID were always slower in various other scenarios. Or if it was just this scenario?

    It would also be interesting to know the absolute numbers. For example maybe most of the test was cached drive C:. Which would mean maybe you are comparing RAM speed with disk speed. (which would look like a 10x difference).


    • #3
      More information

      When using perfmon.msc to look at the data transfer from the drives, I'm seeing around 16MB/sec from the striped drives, which is very good. I need to conduct some additional tests with perfmon.msc to compare it.

      An interesting observation about the graphs though. I run the disc test for 60 seconds. For the first 20 seconds the MB/sec are somewhere between 0 - 50 MB/sec on the scale. At 20 seconds the C: drives quickly ramps up to around 900 MB/sec and stays there. The D: drives remains in the low range for 40 seconds, then ramps up quickly to around 750MB/sec.

      Needless to say, neither the shape of the graphs nor the numbers associated with MB/sec make any sense to me. They do appear to reflect the summary results being reported though. I won't be able to look at this for a couple of days, but I would like to better understand what I am seeing.


      • #4
        The high numbers are a result of disk caching.

        You could avoid this by selecting the uncached option, or increasing the files size, or decreasing the test time.