I'm a relative noob at virtualisation but would like to roll it out in our organisation. The application I'd like to test with needs to be back up pretty quickly after failure... So I'd like to put 2 VMWare servers on the network and have the virtual machines on a disk array so that if one of the hosts goes down I can bring the VM up on the second host. We don't currently have a windows storage server NAS on the network but we do have a Windows Server with a lot of disk space in it.

Also, our internal backups are carried out using DPM 2007 as a result to backup the data it needs to be on a windows machine (I think)

In an ideal world I can share a large array on the windows server and VMWare ESX server use that as a datastore for the VMs.

However, I'm worried that will either be impossible or perform poorly... in which case I guess we use Server 2008 Hyper-V instead of vmware... or is there a better way to provide the redundancy I'm after here?

Finding out my question is meaningless would be as useful as getting an answer because I'm still at the research stage :)

  • The size of the available space is usually not that important. The performance pretty much always comes down to the number of spindles. – Zoredache Aug 5 '10 at 17:38

To make storage available to ESX/ESXi you usually need to use either NFS or iSCSI. You can setup windows to serve NFS using built in software or be an iSCSI target with additional software.

  • There's seems to be lots of setup involved in NFS for Windows... I'm loathe to start extending the AD schema just for this setup but a quick google finds kernsafe with a c$300 iSCSI target software doohickie which looks like it might just cut the mustard. – Paul D'Ambra Aug 5 '10 at 19:20
  • kernsafe "hid" the disk from DPM so we have to backup each guest. No biggie but I'm going to switch that box to FreeNAS to save spending on KernSafe – Paul D'Ambra Sep 16 '10 at 18:31

While the idea of having a redundant server is good, the idea of having a single point of failure at the weakest link in the chain, the storage, is not good. In you scheme, if you have problems with the storage you will be down until it's sorted. Consider having a second copy of the VM drives stored elsewhere if downtime is that important. The data on the secondary storage could be regularly refreshed as part of your backup scheme. While it might not have the concurrency you'd like it will get you back up a lot faster than having to restore from your backup.

  • Keeping in the data on a windows drive so DPM can back it up is probably sufficient in our environment. We've had one drive failure in 4 years and the array carried on chugging away... Although I have now tempted fate :) – Paul D'Ambra Aug 7 '10 at 7:05

Look at Zoredache's answer regarding how to serve up disks for the virtual machines to reside on. Regarding your backup scheme, you have a couple options. You can drop a DPM client directly inside the guest OS, in which case it will ignore the underlying architecture and treat it like another Windows machine to be backed up. You can look here for a way to backup the VMs by tying into VDR. YMMV with that approach (I haven't done it myself, so I can only point you to it for research).

  • we're overdue a test and reassessment of our DR processes which is why we're pushing this move to VM so that it is included when we plan and document recovery. – Paul D'Ambra Aug 7 '10 at 7:04

its a very good idea to use virtualisation together with separate storage-server (san or nas) but

1. i suggest to have two storage-server (one for backups) to have redundancy best are appliances (think about them like your wlan router, that is also a (mostly linux) appliance) you can install them in about 15 minutes and manage them with your browser

2. i would avoid windows (not a good idea to install critical security patches all two weeks with restart)

3. for esxi i suggest zfs-storage appliances with opensolaris or nexenta and napp-it web-gui (easy to use and expand, best data security at all, deduplication, nfs and iscsi for vmware included)

try it at http://www.napp-it.org/pop_en.html easy setup instruction: see napp.it.org

it's free software!

4. if you share storage via nfs then (also mostly faster than iscsi), you can also access your vm's via cifs from windows (easy move or backup them)


  • we're a win-tel network so can't really avoid windows :) Personally I run Mac, Ubuntu Win XP through to 2008 R2 and they all prompt for updates and none require a restart every 2 weeks. You just need to determine the risk to your network from any particular patch. I've had a few colleagues recommend ZFS and if this trial of VMs works out I'd certainly investigate OSS to expand although we get academic licensing from Windows so while we've got internal resource that is mostly Windows based there's a higher cost to the move. – Paul D'Ambra Aug 7 '10 at 6:57

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.