After I wrote the previous blog I realized I left out site replication for your VDI environment. I briefly mentioned that VSC for vSphere can kick off a SnapMirror between sites, but that isn't much help. Site recovery is a huge topic and I could go on for days, so I'll just talk about some basics. Rachel, myself and Trevor Mansell, from Citrix worked on a site replication and recovery paper, TR-3931. It uses Citrix XenDesktop, PVS, XenServer, NetScaler and NetApp SnapMirror, so it might not fit the technologies you're using in your environment, but there's a ton of great disaster recovery (DR) concepts that can be applied to the architecture of your DR scenario.
For example, you need to get your data to another site. In the past this was extremely painful using file based copying technologies which can be slow and usually require dedicated or extra bandwidth between sites. With NetApp SnapMirror, the data is transferred at the block level, only deltas are transferred and if you've deduped on your source side, you don't have to re-inflate the data, send it over and dedupe once again. This is a HUGE benefit!
So what do you copy over? This is going to depend on what you're failover site looks like, but my rule would be whatever you need to make your life easier without utilizing too much storage. DR is hard enough, make it as easy as you can and mirror everything you think you need, you can always tweak things after you've tested a failover.
In the paper we have a pair of active/active sites, so we send Infrastructure and User Data, but leave the Write Cache and Virtual Machine. We leave the vDisk because we have a working copy of it at the other site. If you're architecture is active/passive, I'd copy over Infrastructure, User Data and Virtual Machine. Since the Write Cache is transient data, no need to copy it over.