Rapid Backups: Transitioning from Tape to Disks
If you're still using tape, read on.
- By Greg Shields
I'll never forget the longest day of my career. That day the sun rose, set, rose again, set again and finally rose a third time before I got to sleep. The reason: A major server crash -- or, more specifically, restoring all the data back to that server once its original problem was resolved.
That morning, one of our engineers accidentally changed a configuration on our production SAN full of Microsoft Office documents. His change immediately corrupted a terabyte of corporate data and started a series of events that led to a 72-hour restore. An entire company of thousands of workers sat idle while we watched stacks of tapes do their job.
Trust in Disks
Fortunately, it's a day that needn't ever happen again. That restore could have taken orders of magnitude less time had we replaced our tape-based backup architecture with one that stores backups on disk.
For a segment of IT, disk-based backups represent the first revolutionary shift in most of our careers. The difference is obvious. Reading from and writing to disks is a random operation; one that's much more efficient than the linear-read/linear-write nature of tapes. Today's disks are faster than tapes in nearly any side-by-side comparison. And the capacity of disks has grown to and passed that of most tapes. Low-power, high-density disk solutions now fit into less rack space.
Today's disk-based backup solutions are also more efficient than their tape-based ancestors. Using file system filter drivers, such as the native Block Level Backup Engine Service, among others, the unit of backup shrinks from the individual file to the individual disk block. No longer are you backing up files only because of a time-stamp change, a metric easily spoofed by errant applications. Replacing this ridiculously non-optimized approach is one that backs up, compresses and de-duplicates disk blocks, only when each block has seen a change.
Shift Focus to 'Restores'
But there's a point that's even more important than all of these. Our industry must reprioritize data protection away from its historical focus on "backups." It never should've been there in the first place. Yes, backups are important; you've got to get that data offloaded somewhere else. But backups do you no good if a single accidental configuration change causes an entire company to sit idle for three days. If the focus with tapes is on "backups," then the priority with disks must absolutely be on "restores."
Now is the time to take a hard look at your entire approach to backups. If you lose a critical server -- one whose loss has incredible impact -- can you restore that server in minutes, or will it be more like hours or days? If that same server gets hit by malware, must you set back its clock to "yesterday at 3:00 a.m." or any 15-minute period in the past six months? If you need to restore a document, does that activity require an hour of administrator time or a few minutes? All of these use cases have today been solved by the array of disk-based backup vendors selling Windows products.
When seeking a new solution, know too that you needn't necessarily throw away your now-legacy tape infrastructure. Disk-based backup solutions aren't meant for archival storage. They solve the problem of rapid restore and backup window elimination. Yet a good disk-based backup solution will enable you to "back up the backup" to tape, giving you the off-site rotation and regulation-ready archival your business requires to maintain compliance.
Some inexpensive disks and a new set of backup agents is all your approach needs to turn a 72-hour restore into one that takes just 72 (or fewer) minutes. And trust me, watching the sun rise and rise again before you get to sleep is a situation you never want to repeat.
Greg Shields is Author Evangelist with PluralSight, and is a globally-recognized expert on systems management, virtualization, and cloud technologies. A multiple-year recipient of the Microsoft MVP, VMware vExpert, and Citrix CTP awards, Greg is a contributing editor for Redmond Magazine and Virtualization Review Magazine, and is a frequent speaker at IT conferences worldwide. Reach him on Twitter at @concentratedgreg.