none
Best Storage Configuration for a MD3200 SAN

    Question

  • I have a MD3200 SAS Dual controller storage array with 18 x 2TB disks.  The server this storage will be presented to a Windows 2008 R2 file server.  I'm looking at a few options and would like to determine how they stack up with each other performance wise.

    1. Create two Disk Groups:  9x2TB RAID5 and assign each disk group to a separate controller.

    2. Create one Disk Group:  18x2TB RAID6 and create two Virtual Disks and assign each virtual disk to a separate controller.  

    3. Create one Disk Group:  18x2TB RAID6 and create one Virtual Disk

    Your expertise advice is appreciated,
    Thanks
    Wednesday, July 18, 2012 4:58 PM

Answers

  • It depends on reqirement. For the performance of these options, I think this article provided detailed information:

    http://en.wikipedia.org/wiki/RAID#RAID_10_versus_RAID_5_in_relational_databases

    Please Note: Since the web site is not hosted by Microsoft, the link may change without notice. Microsoft does not guarantee the accuracy of this information.
     

    See:

    Level
    Minimum # of drives** Space Efficiency Fault Tolerance Array Failure Rate*** Read Benefit Write Benefit
    3 1 − 1/n 1 drive n(n−1)r<sup>2</sup> (n−1)X* (n−1)X*
    4 1 − 2/n 2 drives n(n-1)(n-2)r<sup>3</sup> (n−2)X* (n−2)X*

    * Assumes hardware is fast enough to support

    If a database is expected to be strongly write-biased, RAID 5 becomes the more attractive option, since RAID 5 does not suffer from the same write handicap inherent in RAID 10; all spindles in a RAID 5 can be utilized to write simultaneously, whereas only half the members of a RAID 10 can be used. However, for reasons similar to what has eliminated the "read penalty" in RAID 5, the 'write penalty' of RAID 10 has been largely masked by improvements in controller cache efficiency and drive throughput.

     


    TechNet Subscriber Support in forum |If you have any feedback on our support, please contact tnmff@microsoft.com.

    Friday, July 20, 2012 9:19 AM
    Moderator
  • I appreciate the info.  I'm stuck with RAID5 or 6 at the moment to maximize space.  I'd love to go raid 10, but it cost to much at this time.  In either of these cases, it seems that even though more drives should provide higher throughput, that doesn't appear to be the case when that array is presented to windows as one volume.  It appears that the SAN really comes alive when it the # of Outstanding I/O's are high (I/O meter) but it doesn't seem like Windows server really ques things up for the SAN to take advantage of this.  Is this accurate?  I'm just basing this off my experience, but don't have any technical information as to why this is or how to change how windows accesses the storage.  

    No. With a proper stripe size (ordinary application specific I/O is >= of it) with RAID5/6 all spindles are active and perform I/O so you can multiply IOPS.

    -nismo

    Friday, July 27, 2012 8:42 PM

All replies

  • RAID5 is a beast. With 2TB el cheapo SATA drives what you do is begging for troubles. RAID6 is a write performance pig. On your place if you use 7.2K rpm spindles (I assume you want to keep ice cold data, don't you?) I'd stick with RAID10. 

    -nismo

    Wednesday, July 18, 2012 5:39 PM
  • I don't follow your slang.  Can you speak tech for me?
    Wednesday, July 18, 2012 5:42 PM
  • It depends on reqirement. For the performance of these options, I think this article provided detailed information:

    http://en.wikipedia.org/wiki/RAID#RAID_10_versus_RAID_5_in_relational_databases

    Please Note: Since the web site is not hosted by Microsoft, the link may change without notice. Microsoft does not guarantee the accuracy of this information.
     

    See:

    Level
    Minimum # of drives** Space Efficiency Fault Tolerance Array Failure Rate*** Read Benefit Write Benefit
    3 1 − 1/n 1 drive n(n−1)r<sup>2</sup> (n−1)X* (n−1)X*
    4 1 − 2/n 2 drives n(n-1)(n-2)r<sup>3</sup> (n−2)X* (n−2)X*

    * Assumes hardware is fast enough to support

    If a database is expected to be strongly write-biased, RAID 5 becomes the more attractive option, since RAID 5 does not suffer from the same write handicap inherent in RAID 10; all spindles in a RAID 5 can be utilized to write simultaneously, whereas only half the members of a RAID 10 can be used. However, for reasons similar to what has eliminated the "read penalty" in RAID 5, the 'write penalty' of RAID 10 has been largely masked by improvements in controller cache efficiency and drive throughput.

     


    TechNet Subscriber Support in forum |If you have any feedback on our support, please contact tnmff@microsoft.com.

    Friday, July 20, 2012 9:19 AM
    Moderator
  • I don't follow your slang.  Can you speak tech for me?

    Sure.

    Read this one: http://www.miracleas.com/BAARF/RAID5_versus_RAID10.txt

    -nismo

    Friday, July 20, 2012 8:41 PM
  • I appreciate the info.  I'm stuck with RAID5 or 6 at the moment to maximize space.  I'd love to go raid 10, but it cost to much at this time.  In either of these cases, it seems that even though more drives should provide higher throughput, that doesn't appear to be the case when that array is presented to windows as one volume.  It appears that the SAN really comes alive when it the # of Outstanding I/O's are high (I/O meter) but it doesn't seem like Windows server really ques things up for the SAN to take advantage of this.  Is this accurate?  I'm just basing this off my experience, but don't have any technical information as to why this is or how to change how windows accesses the storage.  
    Friday, July 27, 2012 5:31 PM
  • I appreciate the info.  I'm stuck with RAID5 or 6 at the moment to maximize space.  I'd love to go raid 10, but it cost to much at this time.  In either of these cases, it seems that even though more drives should provide higher throughput, that doesn't appear to be the case when that array is presented to windows as one volume.  It appears that the SAN really comes alive when it the # of Outstanding I/O's are high (I/O meter) but it doesn't seem like Windows server really ques things up for the SAN to take advantage of this.  Is this accurate?  I'm just basing this off my experience, but don't have any technical information as to why this is or how to change how windows accesses the storage.  

    No. With a proper stripe size (ordinary application specific I/O is >= of it) with RAID5/6 all spindles are active and perform I/O so you can multiply IOPS.

    -nismo

    Friday, July 27, 2012 8:42 PM