Block Storage

 View Only
Back to discussions
Expand all | Collapse all

Increasing spare protection when using DDP groups on VSP-One-Block

  • 1.  Increasing spare protection when using DDP groups on VSP-One-Block

    Posted 08-01-2025 17:07
    Edited by System 10-17-2025 15:02
    Hitachi F/G and E series block storage appliances 
    have these characteristics:
      
      - Use RAID groups in which the unit of measure is "drive"
        ("drive" could be SAS-SSD, NVMe-SSD, or old school HDD)
        6D+2P refers to a raid group composed of 8 dedicated physical drives.
     
      - Use a spare scheme in which:
          - the unit of measure is "drive"
          - multiple spare "drives" can be defined
          - ANY one of multiple spare drives can be used to bring ANY raid-group back
            to full redundancy. (Thus, a high-value HDP pool built using
            the raid groups can be brought back to full redundancy after multiple
            drive failures ... without immediate human help)
         
     
     
    VSP-One block arrays (we will be installing one soon) use DDP-groups.
    My understanding is that DDP groups used a "chunk" based (not drive based) raid scheme.
    If a DDP group had these parameters (17 drives, 6+2 protection), then chuck-stripes 
    would snake over the 17 drives like this
    ddddddppSddddddpp
    SddddddppSddddddp
    pS.....etc
     
    In a traditional array
    if I had 10 raid groups, I might decide to define 4 spare drives. I could have up to 4 drive failures
    and the RGs backing my HDP pool would get back to full redundancy even if the local field support team was very busy
    and couldn't visit me for a few days. Traditional arrays also provide the possibility of having good luck "save
    the day" ... In the 10 N+2 RG system, after all 4 spares are consumed, a lucky storage admin could lose up to
    20 more drives without pool failure. (an un-lucky admin could see pool failure after 3 additional failures)
     
    I believe that, by default, with DDP (with N+2 protection), arrays can get themselves back to full redundancy 
    after 1 physical drive failure. After 2 or 3 failures the array will operate in degraded mode ... or very degraded mode
    until field support is available.  If a 4th drive failed before field support arrived, 
    its game over. (HDP pool is lost) 
     
    Also, with DDP arrays there is no chance of being very lucky due to a lucky drive failure distribution.
     
    How do I increase my "spare protection" when using DDP?
     
    I'm thinking that basing an HDP pool on multiple DDP groups (which I believe is possible) would make luck possible (depending on failure distribution); but,
    would not provide the same benefit as the ability to define multiple sets of spare chunks.
     
    Thanks
     
    Andy



    ------------------------------
    Andrew Romero
    HDSFC1159
    ------------------------------