Service provider takeaway: Customers who are implementing virtualization will need to optimize their virtual machine backup process. Service providers should use these best practices for efficient virtual machine backup.
The deployment of virtual machines (VMs)
Create snapshots for backup
If you're using VMware, you can use VMware Consolidated Backup (VCB) to create a snapshot of all of the virtual machine disk (VMDK) files. Before creating the snapshot, VCB flushes and suspends all of the writes from the VMs for a few seconds to ensure a consistent image is created. Once the snapshot is created, it is presented to a proxy server, which can then back up all of the VMs hosted by that physical server.
While VCB offers good protection for VMware sites, it's not without flaws. Customers will need to have some type of Fibre Channel (FC) or iSCSI SAN in place, since once the VCB snapshot is created, the proxy server that backs up the VCB will need to access that snapshot over the SAN. Another potential problem with VCBs lies in how they are created. When a VCB snapshot on a VMware server is initiated, the write I/O on all VMs on that physical server is paused while the VMDK file is created; there is currently no workaround for this situation. Microsoft's forthcoming Hyper-V, on the other hand, gives more granular control in creating snapshots since it integrates with the Volume Shadow Copy Service (VSS) found on Microsoft Windows Server; snapshots of individual VMs -- rather than all VMs on the server -- can be created.
Take advantage of deduplication appliances
Using a deduplication backup appliance can help cut down on backup times and minimize data stores. Brian Biles, director of product management at Data Domain, said that the process of backing up to deduplication appliances is similar to backing up to disk or tape. However, there are inherent advantages to a deduplication appliance. Data Domain has found that in many of its customer environments, there are cloned VMs. A deduplication backup appliance can minimize the storage requirements for VMDK files since it recognizes similar blocks of data across these files and stores these blocks of data only once. According to Biles, your clients may get 40-to-1 to 60-to-1 data reduction ratios because of the high levels of redundancy.
Mark Eastman, director of solutions marketing at Quantum, said that the company is developing its DXi deduplication backup appliances to recognize VMDK files in backup streams to reduce backup times and data stores. Quantum has discovered that VMDK files are largely "hollow files," in which much of the capacity assigned to VMs is unused. For instance, if 100 GB is assigned to a VM, only 20 GB might be used. Eastman said Quantum is working toward removing the "hollow" space and then deduplicating the remaining used space.
Be cautious about boutique data protection products
Vizioncore's vRanger Pro is a boutique backup software product often used in customer environments in which VMware was adopted years ago. However, boutique products like this are often too specialized for corporations just dipping their toes into server virtualization, and service providers may need to develop specific internal skill sets to use them. In these circumstances, service providers can help their customers avoid headaches by choosing a widely used backup product rather than one specifically developed for VMware.
Possible choices include Asigra Televaulting, CommVault Systems' Galaxy, EMC Avamar and Symantec's NetBackup PureDisk. Asigra Televaulting uses an agentless architecture that enables discovery of all of the VMs on a VMware server and then deduplicates the data as it backs it up. Galaxy, Avamar and NetBackup Puredisk do the same, but without an agentless architecture. Agent-based architectures require administrators to install agents on each VM, whereas Asigra Televaulting 8.0 can identify VMs on each physical server by communicating with VMware ESX Server and then backing them up.
Symantec's NetBackup 6.5 also offers integration with VMware's VCB. When a backup of this VMDK snapshot is done, most backup software views it as just a file and is unaware that it contains information about data within a specific VM. However, Symantec's NetBackup 6.5 is able to identify VMDK files and back them up as files but also scans the VMDK files and recognizes attributes about them. Using this feature, users can restore specific files within a VM rather than the entire VM.
Using these three best practices, service providers can avoid headaches around virtual machine backup and help protect their customers' data.
About the author
Jerome M. Wendt is the founder and lead analyst of The Datacenter Infrastructure Group. You can find his blog posts at www.dciginc.com.
This was first published in May 2008