The way I see it, there are four major ways to cover off DR in a virtual environment, they all have their pros and cons and cost different amounts of $$$. There are even products that fall into both categories because they are highly configurable and agile, but lets cover the basics of what options there are. Listed in no specific order.

1. Application level replication

This can be broken down into specific replication inside a virtual machine from one vm to another. Things like SQL log shipping, application mirroring, or even the humble robocopy fall into this category. Normally these sort of application replication solutions require detailed runbooks in the event of a disaster to actually recover. Depending on the complexity of the application these can also be quite difficult or time consuming. Costs are normally low, unless a third party product is purchased to actually do the replication.

2. VM level replication

There are a number of vm replication solutions on the market, but the basics of vm replication encompass making sure that the disk and configuration files for the virtual machine reside at the DR site on a regular basis. You could use third party applications like vizioncore or platespin to accomplish this or even get tricky with some copy scripts and VCB. Unless you are using regular replication intervals you could find yourself a fair way behind the your Recovery Point Objective. It needs to be planned carefully and needs to take into consideration the load on the infrastructure and particulary the ability of the link to handle the traffic if need be. Now that the vm is down there, you are going to need a small runbook to explain to any system administrators how to activate these in the event of a disaster. This is also dependant on your choice of application, and some are reasonably priced.

3. A-synchronous storage replication

This is the most commonly used Disaster Recovery strategy for clients that have budget for nice storage configurations. It involves the replication of the physical datastores that the virtual machines live on. The production site remains in read write mode, and the disaster recovery site remains in read only mode. In the event of a disaster you would have to attach some ESX hosts to the replicated luns then manually register the virtual machines. So there is still some process involved here, but there are also ways and means around it. Up until now tricky people had been scripting this recovery process in order to automate it, but guess what VMware's Site Recovery Manager which is coming down the line promises to solve these problems by automating these scripts. Lets hope all the storage vendors get on board.

4. Synchronous replication

The floating datacenter - ahhh the dream. Imaging multiple processing sites linked together with high speed WAN links and hardly any latency. Live storage mirroring between these processing sites with ESX clusters spanning physical locations and virtual machines vmotioning across suburbs. Sounds great doesnt it, well it can be done now and people are out there doing it. The holy grail of disaster recovery, active/active datacenter configurations. Be prepared to open the chequebook though.

I could write about this for hours so if any of you want to know any more by all means email me and I will write some more detail and put up some logical diagrams as well.