We have inherited Hyper-V environment (2008 R2) with SAN storage (HP4500 hardware). Two hosts and two SAN's for redundancy. Whilst I understand virtual machines well, I guess, my knowledge of storage is very weak. I am trying to understand how our MSP engineer designed it and why. That was intro part, here is the main part.

We have 5.32TB of usable space on each SAN, and four volumes on Hyper-V host machines: Volume1, Volume2, Volume3 and recently created Exchange volume for Exchange server. On HP management console I can see according (LUN?) volumes have size of 3TB, 1.79TB, 2TB and 1.07TB. It might look over provisioned, by that is not a main question now.

My question for now is how decision was made to create these volumes and for what reason? Why not create one large volume or a lot of small volumes?

I have tried researching and after few articles, especially this one: https://technet.microsoft.com/en-us/library/ff182320(WS.10).aspx#BKMK_LUNs it makes sense. I do not want to make any assumption and posting my question in case it might be wrong, I do not want to confuse other people who is looking for a similar answer. I will try asking instead.

Dividing storage into few volumes is for performance reason and to prevent disks to be too full or overwritten?

How do I decide on which volume should I store my new VM - I should create all system VHD on one Volume and all data VHD on another?

If yes and if a lot of VMs have a single VHD file then what if one volume will have too many of VHDs and size will grow?

  • 1
    My question for now is how decision was made to create these volumes and for what reason? Why not create one large volume or a lot of small volumes? - Why would you think we would have any idea why your admin set it up this way?
    – joeqwerty
    Sep 8 '15 at 14:50
  • that is why I am trying to find out, whether there is a standard behind it, best practices or is it up to each individual admin?
    – Clear Sky
    Sep 8 '15 at 14:58
  • 1
    There are lots of questions in here, many could not really be answered without significant time "in front" of the hardware. Some of the text also demonstrates a lack of knowledge that you will really need to manage this set up correctly (beyond what you have asked about the storage) You need to hire someone to answer these questions for you, once they have been able to establish your exact requirements
    – Drifter104
    Sep 8 '15 at 15:15

The biggest reasons for splitting it up would be:

  • To use different RAID levels
  • To use different speeds of disks
  • Maybe left over from migrating between different clusters, from before you could just do rolling upgrades to a new version of windows, or some other weird hardware migration reasons

I have a similar setup with a Dell MD3600f SAN, and 3 LUNs, each for a cluster shared volume on my Hyper-V cluster.

One LUN is made of smaller 15k SAS disks in RAID10 - this holds the OS drive for the VMs, and the entirety of smaller VMs. The second is a large RAID6 array of NLSAS drives - lots of storage space, but considerably slower. It holds the data drives for most VMs. The third is a tiny CSV for testing purposes - so I can test things like redirected access without affecting the main ones.

I can't tell you why it's setup that way in YOUR environment, but there may be good reasons for doing so. The first thing to do is look at how the disks are configured - what types of disks are they, what RAID levels, and work from there.

  • 1
    Also, each LUN will have its own disk queue, so in a highly transactional environment, multiple LUNs may increase performance.
    – MDMarra
    Sep 9 '15 at 3:19

To elaborate slightly further on the answer from Grant - I have had previous experience with an HP MSA which granted is not the same model of HP Storage that you are running but I can only imagine that the overall architecture is similar.

We learnt from this particular type of storage, that carving up multiple LUN's allowed us to utilise the active/active controller feature of the MSA so that we can have traffic going through multiple controllers. This meant that we could utilise the cache on both controllers, whilst also having a method of load balancing I/O between the controllers. That is perhaps another architectural reason as to why he created multiple LUNs.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.