By Antony Adshead, UK Bureau Chief
While server virtualization has made life easier in many ways for IT departments, the virtual server backup process has until recently been a difficult and fragmented process.
In fact, while server virtualization went mainstream a few years ago, the virtual server backup process lagged behind; only in the past year has it been integrated into the virtual server environment in a workable way.
What we've seen is an evolution that started by application of the tried and trusted methods of backing up physical servers to the virtual environment, via an awkward two-stage process, and finally reaching a good level of integration between backup and virtual servers with the release of vSphere's vStorage APIs for Data Protection (VADP) in 2009.
This overview will look at the various methods of backing up virtual servers available.
Virtual server backup meets traditional backup
Many virtual server users have dealt -- and still do deal -- with the backup of virtual machine files by pretending they aren't virtual at all. They load the backup software agent onto the virtualized server just as they would have done with a physical server. It works -- but with many agents on many virtual servers on one physical device, the potential for crawl-inducing I/O loads looms large. VMware and Microsoft Hyper-V users often adopted the same approach.
It's the backup equivalent of fighting today's war with the last war's tactics and technology; you may succeed, but it's an approach that could see you suffer losses to attrition or a catastrophic failure.
In answer to these difficulties VMware came up with VMware Consolidated Backup (VCB).
Virtual server backup with VCB
VCB brought some relief from the potential I/O storms of traditional backup on virtual servers, but it was not an elegant solution.
VCB allowed backups to take place at the level of the virtualized server, but its inelegance lay in its requirement to go through a two-stage backup and restore with a discrete disk staging area between the device being backed up and the target.
The process goes something like this: First, the backup application tells VCB it wants to do a backup. This then triggers a Virtual Shadow Copy Service (VSS) snapshot (either of the data or the full virtual machine VMDK volume and data, depending on whether it is a Windows environment or not), and this is exported as a virtual drive image to the staging device. The backup application then takes its backup from that virtual drive image.
Restores of a virtual machine also require two steps. First, the backup software restores the data to the proxy server and then uses VMware vCenter Converter to restore that to the ESX server.
A number of point products have been designed to address the shortcomings of VCB backup, such as Quest Software's vRanger Pro, PHD Virtual Technologies' PHD Virtual Backup for VMware (formerly esXpress) and Veeam Software's Veeam Backup & Replication. These products do VMDK-level full and incremental backups and file-level restores but work in very different ways, so you have to determine what works best for your environment.
vSphere and Hyper-V smooth the virtual server backup process
Last year VMware introduced vSphere and the vStorage APIs for Data Protection, which have made virtual machine backup a much simpler process. Microsoft has meanwhile rolled out Hyper-V's backup architecture, to similar effect.
vStorage APIs for Data Protection allow users to carry out virtual machine backup without copying data to a proxy disk. Now a backup application can talk directly to the VMware VMkernel without scripts or agents mediating the process, and allow incremental or full backups directly from the virtual machine.
Microsoft Hyper-V users only need to ensure the backup product knows it's talking to a Hyper-V server. This is similar to the operation of vStorage APIs, in that Hyper-V virtual machines can be backed up without guest-level backup inside the virtual machine.
The vStorage APIs for Data Protection also include vStorage APIs for multi-pathing, which provide for better storage I/O throughput and storage path failover for a specific storage array; and vStorage APIs for Site Recovery Manager, which enable array-based replication for block and network-attached storage (NAS).
vSphere 4 also introduced a number of other backup-related improvements over VMware Infrastructure 3. These include VMware Data Recovery (VDR), which provides snapshots and data deduplication; granular file-level recovery without a two-stage process; changed block tracking (CBT), which allows the VMkernel to track the changed blocks on a virtual machine's virtual disk to allow only changed blocks to be copied; vCenter Server plug-ins, which provide monitoring and management views of a virtual machine's backup status; direct backup to shared storage; thin provisioning; and data deduplication.
This was first published in October 2010