Tape and Disk Better Together

I have seen a few surveys recently that tape penetration in data centers remains very high, less than 15% of data centers have become tapeless, of course that means that 85% of environments still have tape. In my conversations with IT managers most are planning to keep it. Most see the role of disk in the backup process to augment or at best compliment tape. What's needed then is a way to make tape and disk better together.With each generation tape just gets faster. The challenge is, as you probably know all too well, is that if you can't keep a tape saturated with data, the time it takes to slow the drive down, wait for data and then speed back up can extract a serious performance penalty. Disk is more forgiving. Disks' challenge, although improved with technologies like deduplication and compression, is it is still more expensive than tape. With LTO-5 tape is now about $50 per TB. The most competitive disk systems, even with full deduplication and compression, are typically around $1 to $2 per GB or $1,000 per TB.

The disk price erosion, helped by technology, has certainly made disk a viable short term backup destination and especially because of technology, has make it an option to consider for medium term (~ 7 years) archive. Disk is more forgiving of variable backup performance than tape is, plus it has the perceived advantage of better accessibility because of its file system like nature. Although as we discuss in our article "What is LTFS?" we think that tape as a file system may remove that advantage.

Given these realities can tape and disk be better together? Disk can be simply used as a front end cache to tape. By queuing up large sections of backup on disk first, tape can then be streamed at full speed. This keeps the disk capacity investment down and tape speed up. If handling most recoveries from disk is a goal the cache can be sized up a bit, but typically this is when backup managers will look to a disk system designed specifically at being a backup target, which will also mean adding capabilities like scalability, deduplication and/or compression. The challenge with many of these devices is their integration with tape or lack there of.

Ideally what is needed is an abstraction layer that makes the backup target separate from the software. In some cases the backup software itself can provide the abstraction layer meshing the different backup targets into a single managed infrastructure. Of course that this means selecting only one backup application for the enterprise. For most data centers having only one backup solution sounds nice but is not sustainable, there are almost always a collection of "one-off" data protection processes. Alternatively the abstraction can be done by an appliance allowing multiple backup applications to virtually see whatever device they prefer yet have the appliance manage all the actual back end devices. This would allow a selection of backup targets based on need instead of by application support.

Track us on Twitter: http://twitter.com/storageswiss

Subscribe to our RSS feed.

George Crump is lead analyst of Storage Switzerland, an IT analyst firm focused on the storage and virtualization segments. Find Storage Switzerland's disclosure statement here.