Storage Tiering…

Ok, given the changes to the storage arena I’ve been working on a revised “Tiering system” to incorporate all of the levels of data…importance?

My version of Storage Tiering is (or should be) as follows:

  • Tier-1    – Symmetrix/Replicated – High Performance/Criticial Data
  • Tier-2    – Symmetrix/NonReplicated – High Performance/Non-Criticial Data
  • Tier-3   – Symmetrix/SATA/Replicated – High-Medium Performance/Critical Data
  • Tier-4   – Symmetrix/SATA/NonReplicated – High-Medium Performance/Non-Critical Data
  • Tier-5    – Clariion/FC/Replicated – Medium Performance/Critical Data
  • Tier-6    – Clariion/FC/NonReplicated – Medium Performance/Non-Critical Data
  • Tier-7    – Clariion/SATA/Replicated – Low Performance/Critical Data
  • Tier-8    – Clariion/SATA/NonReplicated – Low Performance/Non-Critical Data
  • Tier-9    – CelerraNAS/Replicated – Network Attached/Critical Data
  • Tier-10  – CelerraNAS/NonReplicated – Network Attached/Non-Criticial Data
  • Tier-11  – Atmos – Network Attached / Low Performance
  • Tier-12  – Centerra (Content Addressable Storage) – Low Performance Archive / Highly Available
  • Tier-13  – Primary Tape-In-Library (Automatic loading on demand via HSM)
  • Tier-14  – Primary Tape-Out-Of-Library (Manual Intervention Required)

“Critical Data” vs. “Non-Critical Data” is simply a matter of how long you can be without the data should a failure or accidental deletion occur.  As all data is available in Tier8/9 storage (in theory).

I’ve also considered using Tier1/Tier1B to describe DMX storage vs. Clariion storage, given that there is a LOT of overlap in performance characteristics these days…

Oh, and iSCSI would be somewhere between 10 and 13….

Any thoughts?


Skip to comment form

  1. Jesse

    I think you’ve a couple of fundamental flaws here. Firstly you are describing your tiers in terms of technology (specific EMC references). I believe you should be discussing tiers in terms of service levels, as technology changes over time and what is tier 1 today may be classed as only achieving tier 2 tomorrow.

    Second, you’ve too many tiers. This leads to fragmentation and wasted resources, negating the cost savings of tiering in the first place.


    1. Chris – My experience is with EMC, so I don’t have much else to go on. If HDS or IBM wants to bring me up to speed on their stuff, I’d be happy to, until then, I’m an EMC guy. 🙂

      That being said – I think my fundimental flaw is I’m making it too complicated.

      What I’m looking for is simply something i can put on a storage request form when an internal customer requests storage. I actually don’t want to give them too many choices, but i do want standards and SLA’s to be in place.

      My earlier reply stated one thing that most people forget, that non-replicated (or asynchronously replicated) data can perform faster than synchronously replciated data, which is usually (in all technologies) the definition of “Tier-1”.

      So do we Tier the storage purely on the basis of performance and put a box next to it for “Replicated”?

      Personally I’m for not giving the end-customer too many choices..the storage admin needs flexibility to do what they think is best…even overriding the customer if necessary. (IE I’ve got a number of windows fileservers on Symmetrix storage, simply because it was the only replication technology available to them at the time – that needed to never happen.)

  2. Hi Jesse,

    A lot of places I see these days have things like Tier1B and Tier3C. With the letters appended at the end referring to things such as shadow copies/timefinder copies etc….

    Also some places distinguish between synchronous and asynchronous replication.

    A few years ago I did some work on a large account that named their tiers after precious metals – platinum, gold silver…… They soon ran out of precious metlas, and rather than resort to things like “tin” and “rusty tin” they started appending letters.

    Im assuming you are also subconsciously pointing out that there are way too many tiers out there. Im sure if you add in things like shadow copies and sync vs async you could get the list to well over 25 😉


  3. I think I’m subconciously forgetting that:

    A: Not everyone uses EMC (although in my little universe they do)

    B: There is a significant amount of overlap these days between tiers and storage systems.

    IE an optimally configured Clariion w/ FC disks can outperform a Symmetrix with a suboptimal SATA configuration.

    So breaking it into simply performance-based groups would give you 7 tiers.

    1 High Performance Replicated
    2 High Performance NonReplicated
    3 Mid Performance Replicated
    4 Mid Performance NonReplicated.
    5 NAS
    6 CAS
    Z Backup

    Though #7 can be sub defined as follows:

    7a DiskIbased
    7b Tape Library Based
    7c Manual Load Tape.

    (We currently have a LOT of 7c storage, so I can’t discount it completely.)

    Then you get into the additional conflict, that “Tier-2” storage will usually perform better than synchronously replicated Tier-1 storage, and how do you sell that to a customer?

    Too many questions for 7am.

    • william bishop on July 11, 2009 at 8:28 am
    • Reply

    That’s not nearly as bad as the original. Of course, once you step outside of EMC, some of those change. There are fast NAS solutions (for instance HDS places a bluearc head in front of their usp, and there is some outrageous connectivity to the box this way)…

    Other than the few exceptions, that list looks a lot better. Why would you not use that list?

    1. Everything changes, but for the moment I can only speak to EMC, because I, as a rule, (and unlike some in the industry) refuse to speak to subjects i’m not familiar with. 😉

      If Hitachi or IBM want’s to pony up with some free training I can speak to them as well, but until then I’m stuck being kind of EMC-Centric. 😉

    • william bishop on July 11, 2009 at 4:58 pm
    • Reply

    I think the first poster’s response of “EMC only” was mainly towards the naming, the dynamics of the arrays aren’t very different at all (I own all three vendor’s arrays). Meaning, if you take out the specific language(model and manufacturer), it all fits.

  4. Totally forgot Tier-0.. (Flash)

  5. How about just breaking it down by application?

    Database/Transactional – VeryBusy (Symmetrix or Clariion High-Performance)
    Filesystem – Busy (Webservices) (Symmetrix or Clariion Medium Performance)
    Filesystem – Not Busy (Archive and such) (Clariion/SATA)

    I think that’s about it. I only throw tape in because where I am there is a LARGE “Tape-On-Demand” system for long-term storage. I hate it, I think there are a million better ways to accomplish what they’re doing, but they’re pretty old-school and fairly set in their ways.

    I also think that with the introduction of 10G Ethernet into the environment (seeing the first switches rolling into the testing lab now) that NAS will split into high/low performance.. Right now the bottleneck is the network so it really doesn’t matter what you put behind it. 😉

    • williamwbishop on July 14, 2009 at 5:35 pm
    • Reply

    I don’t know Jesse….I like tape. It’s portable, cheap, and easy to bring back into systems.

    I detest my centera’s with all of my existence. I spend too much every other year in expansion, I pay maitenance and support, and I pay power and cooling, as well as losing valuable real estate. All so in the unlikely event a radiologist wants to view an old study, it will be a tiny bit faster.

    My goal for this next year or two is to get rid of the bloody thing and go back (where it should never have left) to tape. I’ve got an IBM tape system that will stream 200MB a second to each head….Why not go the cheaper, more portable, and more reliable way? I’ve had EMC working for two years on an ongoing issue with the old gen3 boxes in my setup. They asked me if I wanted to continue maitenance and support for it…My response? “You still haven’t fixed it yet after two years…Why do I need more of that?”

Leave a Reply

Your email address will not be published.