Dynamic Provisioning: Best Practice Guide (unofficial)

By | February 13, 2009

A while ago I put together an HDP/ThP Cheat Sheet come Best Practice Guide for a large UK financial house.  I mentioned it recently on the HDS and HP storage forums and as a result have had several requests for it.

As a result of positive feedback, I have worked on it while at home in the evenings and believe it is now at a good enough standard to make available for anyone to download.

Please note though, I do not currently work for Hitachi, HDS or HP so am therefore not an authority on  the USPV or XP24000.  The content of the document is from a combination of my opinions, experience and knowledge.

The document is a work-in-progress and I will do my best to keep it up to date, but I do have a day job as well.  Oh and a wife, two children and a life 😉

I welcome all feedback and hope that it helps people in their understanding and deployment of Hitachi Dynamic Provisioning / HP Thin Provisioning software.

If anybody would like me to cover any other topics, please contact me and I will do my best to cover them via posts on the blog site.  Also, if anybody, including HDS or HP would like me to write anything for them, or hire me, they are free to contact me.  Im a nice guy and always keen to help where I can 🙂

Nigel

HDP-ThP Best Practices v1.1.pdf

Version 1.2 now available – here

4 thoughts on “Dynamic Provisioning: Best Practice Guide (unofficial)

  1. John Harker

    Interesting article.  Working for HDS, it’s great to see documents like this coming out of the user community.
     
    Reading it I had a few comments.  
     
    First, with certain exceptions our best practices for dynamic provisioning pools is that they be kept homogeneous.  We don’t think attempting to improve performance through maximizing spindles in the pool using multiple drive types is likely to work.  Essentially the Pool must be rated at the weakest point – effectively the bottle neck.  In general adding fast fibre drives to a SATA pool does not make the pool faster or better it just wastes the fast drives.  .
     
    One exception to keeping pools homogeneous is that a new generation HDD might be added to a pool. For instance adding more 300GB to a 146GB pool (after 146 is no longer available).  But you should keep the same RAID and RPM.
     
    You were spot on with the suggestion to just put one DP-VOL in each V-VOL group so as to be able to expand it later.
     
    With regard to multiple pools, although there is the argument that the larger you make a pool the more you will benefit from natural load balancing and hot spot reduction, there is a counter argument that there is almost always a need for more than one tier of storage or the need to isolate individual workloads.  Sometimes the decision is driven by operational or management considerations, sometimes by technical storage design considerations, but in most cases users wind up with multiple pools representing multiple classes of storage.   
     
    Later, in the section on items to be considered if there is a requirement to utilize the oversubscription feature of HDP, you could mention:
    –         After migrating a ‘thick’ volume to a virtual volume it is good practice to perform a ‘zero data discard’ operation, something better known as Zero Page Reclaim.  Often quite a bit of physical storage can be freed.  This is a new feature and requires USP microcode  60-04-04 or higher and an updated Release of Storage Navigator associated with this microcode release.
    –         Extra benefits can be obtained with replication where the PVOL and SVOL are both thin – only the allocated portions of the thin volumes are replicated which saves on overhead and bandwidth.
     
    A colleague read it and marked it up as well, if you mail me separately I will send you his comments too.

  2. sruby

    >>-         After migrating a ‘thick’ volume to a virtual volume it is good practice >>to perform a ‘zero data discard’ operation, something better known as >>Zero Page Reclaim.  Often quite a bit of physical storage can be freed. >> This is a new feature and requires USP microcode  60-04-04 or higher >>and an updated Release of Storage Navigator associated with this >>microcode release.
    This is available on the USP or USP-V?
     
     

Leave a Reply

Your email address will not be published. Required fields are marked *


*

You can add images to your comment by clicking here.