1

I am planning on deploying a NAS/iSCSI target cluster of two Dell NX3100 servers running Windows Server 2008 R2.

Here are some of the environment details currently:

  • I have two office locations on opposite sides of the US that will host one of these NAS/iSCSI target servers each.
  • The locations will also have XenServer hosts at each location. Each site has a DC/DNS XenServer VM, amongst others.

My questions are as follows:

1.) Is a point-to-point connection or site-to-site VPN more suited for this setup? Maybe a third connectivity solution?

2.) I plan on using half of the array on the NX3100 servers as iSCSI targets for location-respective XenServer VHDs. I mean to run the VMs off of their local NX3100, but backup/snapshot to the off-site NX3100 for disaster recovery.

3.) Are these ideas silly/unfeasible? What are some foreseeable problems or misunderstandings I may have with this environment?

BeepBoop
  • 283
  • 2
  • 10

1 Answers1

2

Definitely split the two locations into AD Sites and set your replication schedules accordingly, that should help with the cross-site connectivity and coordination. I'd strongly recommend using a VPN for this traffic. Alternately, a point-to-point using IPSec between the Sites would be my second choice.

For backup/snapshot, definitely keep your sync times in mind when scheduling your AD syncs and plan to avoid big AD changes during the time periods when you're doing your cross-site backups. The backup/snapshot stuff will be the most sensitive to bandwidth conditions between the two sites. Definitely use an asynchronous bu/snap method, those distances are f-a-r too large for synchronous ones. Only you can predict what your likely transfer needs will be during this, so test and watch things.

The one red-flag I see is that your snap/bu may not complete in time to be useful. Keep a close eye on those.

sysadmin1138
  • 131,083
  • 18
  • 173
  • 296
  • Thanks for the great answer! I will remember this for my AD. I believe DFSR in server 2008 supports asynchronous I/Os; thanks for alerting me to the potential issue there. The snaps may be a bigger issue, but I'm really only relying on them for true disaster. – BeepBoop Jul 20 '11 at 17:40