Going virtual? Don’t get physical

When it comes to backing up—and restoring—virtual servers, don’t rely on the same system used to back up physical servers.

virtual-servers-backupAs college IT departments look to wring as much value as possible from their budgets and infrastructure, server virtualization has become almost de rigueur. While the benefits of virtualization are now well documented, maintenance of this new environment is less understood—and comes with its own set of challenges.

Central among these is the question of how virtualized servers can be efficiently backed up and restored.

Among IT departments that have recently made the transition to a virtualized environment, a common approach—or mistake, as some would see it—is to use the same backup process for their virtual machines (VMs) as for their physical servers.

“A lot of educational institutions have gone from a largely physical data center to a highly virtualized one, but they haven’t changed their backup processes,” confirmed Doug Hazelman, vice president of product strategy for Veeam, a Swiss company that focuses on management solutions for modern data centers. “By using agents and those types of things, they are still using backup processes designed for the physical infrastructure. Because of that, they’re seeing their backup times increase, even as customer demands are also increasing in terms of availability.”

This is exactly what Red River College in Manitoba experienced when it first virtualized the server setup at its three main Winnipeg campuses. “With physical servers, backup had been fairly easy because we had tape drives on most of our devices,” said Darren Toews, acting senior technical analyst. “We didn’t have any dedicated storage devices to host disk-to-disk backups, so we started off by just running Windows backups on the VMs themselves, backing those up to a file share on a server and then backing that up to tape. It was an inelegant method.”

While the system worked, its limitations became quickly apparent as the number of VMs grew. “The backup process created a lot of network traffic, and the backup windows were always really, really tight,” recalled Toews.

Recognizing that its approach was unsustainable, Red River searched for a way to automate and shorten the backup process—and to provide rapid recovery of failed VMs.

(Next page: The power of snapshots and reporting tools)

After trying a couple of third-party solutions, the school settled on Veeam’s Backup and Replication tool.

The Veeam product provides VM backup for VMware vSphere and Microsoft Hyper-V virtual environments. Unlike traditional backup agents that go inside the server and the VM, back up all the individual files, and then stream those through a media server, the Veeam product takes what Hazelman calls “an image-level backup of a VM. We tell the hypervisor to take a snapshot. Then we can either grab the entire snapshot or just grab the changes at the disk block level from the last time it was backed up.”

Not only does this approach speed up the backup process, but it’s married to a reporting tool that provides status updates. Every morning now, Toews receives an e-mail report detailing the status of the previous night’s backup. “Each day, I know what my recoverability state is,” said Toews, who operates six Veeam backup servers. “From there, I can figure out why a backup may have failed and make sure that tonight’s backup is going to work.”

As any system administrator will explain, though, backups are only the beginning of the story. “Backing things up is great, but if you can’t recover your backups they’re not really doing anything for you,” explained Toews. “Recovery is the key.”

Before Red River implemented Backup and Replication, the IT team had to test backups manually to verify whether they were good. Considering that most IT staffers spend the workday with their hair on fire, this was a problem. “It’s one of those tasks that typically seem to get pushed to the side because you have so much operational stuff to get through on a day-to-day basis,” said Toews.

This was why Red River was so enamored of the SureBackup feature of Veeam’s tool. When activated, SureBackup powers on the backups and runs a series of diagnostic tests. “You can have the system run tests to see if it can establish a connection on a port that should be listening for a mail server, for example,” explained Toews. “Now, not only do you know that your backup powered on, but it is actually listening on that port. It tells you that your mail application is working.”

Because SureBackup runs on the VMware host, it does impact performance slightly, so Red River doesn’t verify every VM backup daily. “We do key services at least once a week, sometimes twice, to make sure they’re working,” said Toews, who would like to reach the point where the backups are verified more frequently. “Knowing that that backup is good is what could save you in the long run. You don’t want to find out that it didn’t work when you need that backup.”

With IT services assuming an ever-greater role in college life and constituents expecting 24/7 availability, recovery also needs to be quick. “The key is being able to recover in a variety of ways,” advised Toews, noting that schools need the flexibility to recover individual files, an entire virtual machine, or individual drives on a virtual machine. To streamline file-level recovery, for example, the Veeam product utilizes different wizards that require just a click or two, as well as a web interface that allows users to search for files and then recover them.

In the event of a VM failure, Veeam’s Instant VM Recovery allows IT to spin up a backup as if it were a VM and actually run the production server off the backup while it restores in the background. According to Hazelman, the whole process takes under 10 minutes. “The security of being able to run off last night’s backup with nobody noticing much of a difference is key,” said Toews. “It let’s you hit those hard-to-hit recovery-time and point objectives.”

Regardless of whether the servers are virtual or physical, Toews emphasized how important it is for schools still to abide by the 3-2-1 rule for backups: “You want to have three copies of your data on two different types of media, one of which is offsite.”

Andrew Barbour is an editorial freelancer with eCampus News.