Welcome to my first post here on Virtualization Review. Bruce Hoard was kind enough to invite me to be one of the first bloggers here in this corner of the Internet, and I jumped at the opportunity to share some of my thoughts on virtualization with a wider audience. I'm hoping this series will be informative, but also humorous at times.
We all know that with virtualization, there are a lot of changes that need to be made in the datac enter. Besides the benefits of server consolidation, power reduction and high availability, there's also a great opportunity with virtualization to change your backup strategy. Gone are the days of requiring backup agents on every server; virtualization makes image-level backup a reality. But as with all new technologies, change is sometimes difficult to embrace, especially when it comes to what runs the business: the data.
For the past few decades, tape has ruled when it comes to backup. Tape has also locked us into a particular way of thinking about backup and how it should be performed. I'm not saying tape is bad -- it has its uses for long-term archiving -- but should it rule how your backups are performed? Tape is a linear technology – you start at the beginning and write to the end. This means there's no "random access," so you can't go back and insert changes very easily. As a result, backup technology has been very slow to progress out of these full, incremental and differential methodologies. Thanks to tape, everyone thinks the best way to backup is to perform a full backup once a week and then do a daily incremental or differential backup until it’s time to do the next full backup. This makes sense with tape, because you can't go back and "change" the full backup once it's on tape. Also, since we don't want to load 365 tapes to do a full recovery, the industry has accepted that weekly full backups are "the norm.”
How Can Virtualization Change Backup?
With virtualization, servers (VMs) are nothing but files on disk. This makes it very easy to backup the entire VM as an image instead of installing agents on the server and backing up just the data. Recovery is also simplified because the entire VM can be restored onto the hypervisor. There’s no need to first build a server, install the agent and then recover the data. With these changes, is it also possible to take a second look at "the normal" backup procedures? I think so. When you look at backup methods, full backup is always slowest, differential is faster, and incremental is fastest. Why not always do incremental backups if it's the fastest approach? The answer is easy: Tape (and that's the way it's always been done)!
What if you were writing your backup data to disk rather than tape? Disk is random access, you can go backwards and forwards and quickly change any "data" that's stored on disk. After all, this is why modern computers have disk drives and not tape drives. When writing the backup data to disk, the promise of synthetic backup finally becomes a reality. Synthetic backup technology is not new, of course, but the way in which vendors are approaching it with disk-based backups today is truly exciting. With synthetic backups, you perform a full backup just once and then make incremental backups forever after that. If you're still thinking in terms of tape, this sounds scary. But when you consider that you can "update" the full recovery after each incremental backup (because it's stored on disk) you start to see the advantages. Synthetic backup offers the fastest backup (always incremental) as well as the fastest restore (your full recovery is always the most current).
Don't Shy Away from Change
I have seen some comments recently that try to scare people away from synthetic backups. But this is because they don't understand the technology. When done properly, synthetic backups offer all the benefits of the normal full/differential/incremental approaches without the pain of performing weekly full backups. As one of my friends put it the other day, why would you backup a file every week that hasn't changed in three years? With tape you have no choice. With disk you do. As virtualization continues to force changes in the data center, isn’t it time you took another look at your backup strategy?
Posted by Doug Hazelman on 05/19/2010 at 12:49 PM2 comments