IT teams that have gone big on virtualization can take advantage of advances that simplify the process of backing up virtual machines.
VMs are composed of fixed configuration and disk files, so it's easy to grab a copy of a machine's resource profile, snag the associated disk files, and spin it up anywhere, anytime. And VMware and Microsoft have improved this functionality.
Do you store your VM images on a SAN? Want to back up your machines while they're still running? Snapshot-compatible backup applications that leverage vendor APIs to do the heavy lifting are available for the major hypervisors. They make API calls to take snapshots of VMs prior to starting the backup process. Changes are redirected to a change file instead of the original volume. The machine is backed up, even if clients are using it, and the snapshot is destroyed.
VMware has performed backups this way since ESX 2.0, but it required the use of a VMware Consolidated Backup proxy server. This meant IT had to deal with tedious configuration requirements, and the underlying storage system was saddled with a significant amount of read I/O during backups.
Users griped, so VMware has eliminated VCB in favor of an integrated API called vStorage APIs for Data Protection. VADP is invoked directly by a backup application and is a configuration-free feature of vSphere.
Third-party virtualization backup providers such as Veeam have taken full advantage of VADP and its Microsoft equivalent, Volume Shadow Copy Service, to deliver low-cost, highly effective, VM-level backup at a reasonable price.
Virtualization APIs can also grab flat-file backups from guest virtual machines, and products that already excel at flat-file backups can leverage the APIs to do so better and with a smaller footprint. But most exciting is underlying storage's involvement in the backup process.
The latest storage APIs from VMware, and to a lesser extent Microsoft, use the horsepower of the storage arrays to make copies of data to be backed up. This approach works provided the storage array has enough spare capacity to accommodate the I/O. Doing the backup itself requires an API-compatible backup product for VMware. Acronis, CommVault, EMC, Hewlett-Packard, Quest, Symantec, and Syncsort are API certified. Using hardware-assisted copy requires a SAN that's compatible with vStorage API for Array Integration. Three vendors have VAAI-ready SANs, according to the latest VMware hardware compatibility list: FalconStor, LeftHand, and HP.
Other advances are transforming backup. For instance, imagine running traditional network- or agent-based backup over the internal hypervisor instead of across conventional network links. VMware's Virtual Machine Communication Interface allows machines on the same host to read data from one another at the speed of the internal machine bus. It's fast, particularly for large transfers or where the data is in running memory. One example is SQL Server, which is designed to keep as much data as possible in memory for faster reads and queries.
Note that you'll need to upgrade to vSphere 4 or later to take advantage of these APIs.
Jake McTigue is president of McTigue Analytics.