Topic
6 replies Latest Post - ‏2013-12-16T11:45:04Z by eJones
hans72
hans72
4 Posts
ACCEPTED ANSWER

Pinned topic v5000 - recommendation for EasyTier setup

‏2013-12-09T14:12:18Z |

Hi,

I consider replacing old DS with V5000 and got a novice question about EasyTier. I understand with EasyTier data are moved between different MDisks, therefore (considering reliability of the SSDs) it is recommended to use RAID protection for the SSD MDisk.

Is it possible, though,  to use HDD as a hot-spare for SSD MDisk? I am aware of the performance implications, just wonder if this would be possible?

Also, is there any danger in using RAID 5 for SSD pool? Without moving parts in SSD, the difference in write speed should only be the time required to calculate the checksum and surely V5000 should have enough power to cope with that? Or should one stick to RAID1/RAID10?

Thanks in advance!

Hans

  • AndersLorensen
    AndersLorensen
    139 Posts
    ACCEPTED ANSWER

    Re: v5000 - recommendation for EasyTier setup

    ‏2013-12-12T16:10:12Z  in response to hans72

    You should always raid-protect your SSD's in a Storwize.

    Whenever you use Raid1, 10, 5 or 6 depends on your wallet, your requirements etc. I normally suggest Raid5 to people if they have 3+ SSD's. But if you expect a TON of IOPS (50k+) on them, you might want to consider other raid types.

    I've yet to configure a SSD as a hotspare. Their current pricing and reliability just dont make space and sense for a hotspare. And no. You cannot use a HHD as a hotspare for a SSD. If you Loose a SSD in a MDISK, and it looks like its gonna take a long time for IBM Support to replace it (or whoever replaces your SSD's), you can remove the MDISK from the pool, and migrate the data away to other MDISKs in the pool.

     

    /Anders

    • chriscanto
      chriscanto
      230 Posts
      ACCEPTED ANSWER

      Re: v5000 - recommendation for EasyTier setup

      ‏2013-12-13T08:59:19Z  in response to AndersLorensen

      Just to correct one thing Anders said, the Storwize family systems *will* use an HDD as a spare for an SSD.

      The system terms any drive with sufficient capacity but inferior technology a "Capacity match" and is a legitimate candidate for sparing if no other performance match exists.

      Performance might (will?) obviously be impacted in your SSD array if it took a nearline spare, but its better than data loss due to double failure.  As Anders says, you can remove the degraded array mdisk from the pool if it looks like a replacement SSD is going to take some time to happen.
       

      • AndersLorensen
        AndersLorensen
        139 Posts
        ACCEPTED ANSWER

        Re: v5000 - recommendation for EasyTier setup

        ‏2013-12-13T09:18:52Z  in response to chriscanto

        Interesting. I've never seen a SSD actually fail, so my "wisdom" came from the GUI.

         

        When I look on one of our V7000s, when I take propertires on a SSD MDISK, Under Member drives tap, it says "# of Spare Drives: 0", when I have 2 SAS disks with enough capacity to act as a "capacity match". On my SAS MDisks, it says "# of Spare Drives: 2".

        I guess the GUI just dont tell the Whole story and only shows Exact match spares...

        /Anders

        • chriscanto
          chriscanto
          230 Posts
          ACCEPTED ANSWER

          Re: v5000 - recommendation for EasyTier setup

          ‏2013-12-13T10:09:53Z  in response to AndersLorensen

          I think you're right.  Clicking on the 'Show Details' checkbox on the mdisk details dialog does make a "Suitability" field appear next to the "# Spare Drives" and the value looks to be fixed to "Exact Match".

      • Mach_v
        Mach_v
        7 Posts
        ACCEPTED ANSWER

        Re: v5000 - recommendation for EasyTier setup

        ‏2013-12-13T20:29:38Z  in response to chriscanto

        We have test v3700 (but i'm think all storwize/svc family have same code) at one of our customer, which prefer have request in one of pre-production-ready test to check how system will go through hdd/ssd failure, and with 2 hot-spare at chassis we test remove one hdd and one ssd from it, with return "failed" (removed) disk back after 10-25 min (when Microsoft SCOM finaly seen error at storage).

        I'm too wonder when system have get one of hdd spare disk at degraded ssd Raid1 mdisk, and automatically copy it back when ssd been returned at chassis (and at ssd raid1 mdisk there is spare goal set to Zero, to create mdisk with only 2 ssd).

        This moment not so clear at documentation|redbook.

      • eJones
        eJones
        5 Posts
        ACCEPTED ANSWER

        Re: v5000 - recommendation for EasyTier setup

        ‏2013-12-16T11:45:04Z  in response to chriscanto

        We hit this issue last summer with our V7kU system too.

        >mkarray -level raid10 -strip 128 -sparegoal 1 ...

        This will work, and you can use a non-ssd device for sparing if it meets the capacity requirements.

        But the system will still log Error:1690 alert messages, etc. when it detects this.

        We opened a ticket with IBM, and proposed to reduce this to only Warning level, but Development disagreed.

        So we ended up having to get extra SSDs for spare pools, in order to avoid the alert noise cluttering the logs.