It’s about the restores.

By Juan Orlandini
5/24/2011

Back in the day, I was a sysadmin at a large southeastern university. My challenges there were not too dissimilar from "the real world" outside of academia. This was especially true in backups. Back then, the college that I worked for had limited budgets for commercial software. We wrote most of what we needed, including backup scripts. I was really proud of the backup script I'd written for our UNIX boxes. It was very complete, could do some really cool things (for backup software), and it was "mine." Interestingly, the Dean never once came to the data center to give me a high five for doing backups.

Not once.

But he sure did care when I couldn't do a restore.

The lesson I learned from that was it really wasn't about the backups. It was all about the restores. (Oh and don't write your own backup scripts.)

As an industry, we have learned this the hard way. For years, our customers focused on the speed of backups. In reality, the focus should have been on the speed of restores. Recently, this has become a much easier thing to do because we are moving from using tape drives for "backups" to using storage arrays' built-in data protection technologies. Snapshots, mirrors, and replicas are increasingly used as our primary "recovery" technology. Turns out this is a "good thing" − except for one glitch. Array vendors are not recovery experts. They also do not tend to "get" all of the stuff you have to do to protect applications well. Just as bad, they absolutely don't get the concept of a catalog. If you use snapshots for protection, you'll know this to be true immediately. There's nothing in any storage vendor's arsenal that can tell you where you have each copy of every file across all the snapshots you have taken.

It's like having all the backups in the world, but not really knowing where stuff is. So, we are forced to manually hunt and peck for the data that we are asked to restore.

That's the bad news. The good news is that backup software is evolving. It's going from the backup-centric world to the recovery-centric world. There's much that's happening there, but one of the coolest things is that soon you will see backup software that integrates with array-based technologies. This integration will let backup admins control, schedule, and monitor the snapshots, mirrors, and replicas that are used for data protection. They will do all of this and catalog what each of those snaps, mirrors, etc. contain. Very soon, you will be able to go to your backup software and ask it to find the data you are looking for and have it automatically figure out if it's on a snap, mirror, replica, backup to disk, or tape. The software will know which copy is the most recent and desirable and automatically use it to recover your data.

Can't. Wait.