Virtual machines are increasingly the workhorses of the enterprise data center, and no backup strategy is complete without a plan to protect them.
While VMs are, at heart, simply a collection of disk files combined with a configuration file, there are some challenges. First, VM disk files tend to be huge. While having an exact copy of every machine on a backup medium is a fantastic way to recover from a bare-metal standpoint, you have to consider the amount of storage required. And, given the 24/7 nature of today's data centers, we have to back up our VMs while they run and serve clients. Regardless of the hypervisor platform in use, this is accomplished via the creation of a VM snapshot. When a snapshot is made of a VM, the hypervisor stops writing to its existing disk file and creates a new disk file to write changes to. If the machine is live, it also saves the contents of running memory to a separate file. The backup software can copy the snapshot while allowing the VM to continue operating.
In the normal course of a disk-level backup managed by a third-party product, the snapshot will be taken immediately prior to the backup and deleted immediately thereafter. This process requires that any backup product you use be able to do three things: communicate with the hypervisor before, during, and after the backup to request the creation and deletion of snapshots; gain access to the raw VM disk files; and quickly and efficiently move data from the virtual infrastructure out to a storage medium without disrupting the current workload.
While snapshots are useful, IT teams can run into problems if they're not careful about management. Time and time again we've seen administrators use snapshots as a quasi-backup instead of what they're intended to be--a temporary safety net for storing a copy of the VM disk files during the actual backup process.
Now, there are a few reasons IT might leave copies of snapshots on the production server. For example, they can be a quick-and-dirty option to revert to a known-good state if a problem arises on a VM. The problem is that every snapshot consumes storage space. If multiple snapshots are saved, the VM risks using up its alloted storage and then failing.
Download the Feb. 7, 2011 issue of InformationWeek