Re: IBM FastT vs. EMC Clarion

From: Jim Richard (JRichard_at_SCIQUEST.COM)
Date: 03/25/05

  • Next message: Vincent D'Antonio: "Re: IBM FastT vs. EMC Clarion" 
    Date:         Thu, 24 Mar 2005 23:25:50 -0500
To: aix-l@Princeton.EDU

    I have a fastt600 running for a year now flawlessly... My only gripe is that
    it was advertised as being raid 10 aka (1+0) capable... And technically it's
    not. To my mind raid 10 is Mirror the disks across pairs then stripe the
    data across the mirrors. IBM builds the stripe on half the disks then
    mirrors the entire stripe onto the other half of the physical disks.

    This has a number of negative reliability, and recovery ramifications :

            Reliability:

            In my idea of a raid 10 ( and most other vendors' that I've dealt
    with ) you would have to loose 2 individual disks that are directly
    paired together to loose the array.
            So say you have a disk array with 10 drives, this would be comprised
    of
            5 pairs.
            You would have to loose both disks of any single pair for the array
    to fail. So I can actually have up to 5
            disks of the 10 units fail before the array fails (very best case),
            so long as no two failed disks were paired together.

            In IBM's idea of a raid 10 you have an array comprised of 5 drives
    striped ( raid 0) then the raid 0 is mirrored
            to another 5 drives (raid 0+1).
            In this case if you loose 1 drive on one stripe then any other drive
    on the other stripe the
            array is toast. The only way to lose 5 disks and have the array
    survive is if all 5 failed disks are in the
            same stripe. Of course the only likely scenario for this is if your
    stripes are in two separate disk enclosures, and
            you lose an enclosure. Which is a good idea in either case, in my
    idea of a raid 10 no paired disks would reside in
            the same enclosure.

            Recovery:

            In my idea of a raid 10 described above, if you loose a drive, when
    you replace it only the one drive has
            to be re-mirrored.

            In IBM's idea of a raid 10, the entire 5 drive stripe has to be
    re-mirrored...5x I/O, CPU, contention on the
            switch .... Etc.

    To be fair you can mitigate most of the reliability risk with proper use of
    hot spares and diligent monitoring, which should be the norm in any large
    disk deployment SAN, SCSI, or anything else. Unfortunately the recovery
    issue is still a problem for large arrays.

    Unfortunately for me the decision to purchase these units) was based on high
    level marketing material and discounts... I only found all this out after
    slogging through the technical docs ( and this little detail is buried way
    down in them) after the equipment was already on order. A typical case of
    buyer beware.

    Other then that the stuff is pretty nice. If you're not planning on running
    raid 10 (1+0) you should be fine. If you are going to run mirrored you'll be
    fine as long as you follow best practices and understand the issues.

    Jim

    -----Original Message-----
    From: glh@DAIRYNET.COM [mailto:glh@DAIRYNET.COM]
    Sent: Thursday, March 24, 2005 12:31 PM
    To: aix-l@Princeton.EDU
    Subject: IBM FastT vs. EMC Clarion

    We are in the process of selecting our first SAN for our environment -
    approximately 10 AIX servers and 3 Windows servers. We've narrowed our
    choices down to either an IBM DS4300 Turbo (old FastT) or the EMC CLARiion
    CX500. For those of you who may have worked with either (or both) of these
    products, what is your overall opinion of the product? Would you buy that
    product again? Any positive or negative comments would be greatly
    appreciated. Thanks.

  • Next message: Vincent D'Antonio: "Re: IBM FastT vs. EMC Clarion"

    Relevant Pages

    • Re: xRAID disks....
      ... For mirrors it can actually be a big win with hardware support. ... Typical bandwidth for the northbridge/southbridge connection is 1-2 GB/s ... RAID 10 is nice, but it requires more disks then RAID5. ...
      (freebsd-questions)
    • Re: RAID 10-LUN Question
      ... My thoughts are take most even number of possible disks(leaving private ... system partition area) and create the raid groups and kick a separate LUNS ... drives, 12 drives, etc. Depending on how 24/7 self-managed you want the ... Stripe size effects performance but also storage ...
      (freebsd-questions)
    • Re: RAID and Cluster!!
      ... By RAID 1+0 do you mean a stripe of mirrors or a mirror ... A stripe of mirrors gives much greater redundancy than the other ... doesn't matter if it is a cluster or a standalone SQL Server, ...
      (microsoft.public.sqlserver.clustering)
    • How to RAID1+0 with Solaris 9 SVM
      ... I first tried building 14 mirrors, each made up of two partitions, one ... SVM said, no no, cannot build stripe out of metadevices! ... Solaris Volume Manager supports both RAID 1+0 (which is like having ...
      (SunManagers)
    • Re:
      ... Is using LVM2 for striping safer than raid 0? ... RAID0 means striping. ... don't put it on a stripe, put it on a mirror: ... dividing the MTBF by the number of disks. ...
      (Ubuntu)