As we discussed in a recent webinar, “Why Tape is Making a Comeback“, data centers of all sizes are seeing the wisdom of renewing the role of tape and tape libraries in their backup and archiving environments. With the release of LTO 6, the cost per gigabyte, combined with the increased performance and reliability, tape can no longer be ignored. Moreover the inclusion of LTFS to LTO 5 and 6 addresses tape’s previous interchange challenge. While tape has mostly re-emerged as the ultimate archive, data centers would be well advised to consider expanding its role in traditional data backup workloads.
Tape was once the king of the hill of the data backup market but was dethroned, in large part, by sophisticated disk backup deduplication appliances which drove down the per GB cost of disk just enough to warrant displacing tape in many environments. With that technology in place, the pendulum swung to disk being the primary, and in many cases, the only backup hardware an organization had.
As is the case in most big pendulum swings, the swing to disk backup has probably gone too far. A medium to large enterprise that needs to retain dozens of TB’s of data backup for years or decades, is going to have a very expensive backup infrastructure to power and cool. Pulling the pendulum back a little and allowing tape to fulfill a larger role in the data protection process may be a safer and certainly a more cost effective long term strategy.
Know Your Restores and They Will Set You Free
There are two types of restores. The first is the kind where the pressure is on, you have the proverbial gun to your head. The data must come back on the first attempt and that recovery must be lightning fast. For critical applications this likely means that the cluster or replication has failed for some reason. In the case of a “rolling disaster”, locally corrupted data is replicated to the secondary site and corrupts the copy on that side as well. When this occurs, backup administrators need to revert to a point in time backup copy.
In this particular situation, disk backup plays a key role. More than likely, only a small amount of data needs to be recovered and with it residing on disk, it can be accessed very quickly. In almost all cases, the data being recovered is the most recent known good copy. On rare occasions, it may be a version or two back. Frankly, I have yet to see a situation where a copy that is more than a few days old is needed “instantly”.
In almost every circumstance, a need for data that is outside of the “last known good copy” window is generated by a specific information request and in this instance, you have ample time (hours even days) to respond to the recovery request. Purists will claim that this data should be in an archive, but for many data centers, the reality is it is more than likely contained in the backup data set. Our position is that regardless of the underlying process, backup or archive, it should be on tape.
Giving tape a more prominent role in the backup infrastructure does not necessarily mean the outright replacement of disk. Instead, it merely keeps the investment in disk and its associated electrical, power and cooling costs much more manageable. It also sets the stage for a powerful archive strategy to make the data purists happy.
DataSpan is a client of Storage Switzerland