All backup procedures have contributed to some degree to easing the burden of backing up data. All of them, however, have done nothing to solve the Big Problem facing backup administrators.
For the last 40 years backup procedures for host computers and network servers have remained substantially the same: use some tape rotation scheme such as Grandfather-Father-Son (GFS) to save data to tape.
In 1997 IBM introduced the Virtual Tape Library (VTL) and since then numerous disk-to-disk, disk-to-disk-to-tape and disk-to-disk-to-cloud backup solutions have appeared. All of these approaches have contributed to some degree to easing the burden of backing up data. All of them, however, have done nothing to solve the Big Problem facing backup administrators.
At last week’s Tape Summit, a conclave of tape vendors and industry analysts held in Henderson, Nev., this point was brought home by a backup administrator from a Fortune 50 company. He described his trials and tribulations of backing up many terabytes of data and getting that data stored off site in a 24-hour period. He determined that the quickest way to get his data off-site was to ‘hand it to a guy with a truck,’ which is much faster than cloud storage if you have a lot of data.
Although some updates of his tape technology might have added marginal performance (he was using LTO-3 and LTO-4 instead of the newer and somewhat faster LTO-5), it was apparent to the attendees that his backup approach was unsustainable, as he was nearing the point at which he could no longer meet his backup window. Unfortunately, his situation is not unique.
So what is the problem and the solution? The problem is something I like to call TMDD, or Too Much Damn Data. The old model of backing up all of the data on our hosts and servers doesn’t work when you are drowning in a sea of data. (By the way, the size of the data sea is rapidly expanding.) This issue won’t be solved by swapping out GFS for image backup, continuous data protection or some other method, although they definitely have their merits.
Our current solutions for backing up data are much like rearranging the deck chairs on the sinking Titanic. The deck looks nicer but the ship is still going down. The solution will require extensive study of the data itself and establishing management methods for different data types. It will require a good understanding of the full life-cycle of each data type, from creation to backup to archiving to destruction.
In other words, it will require a complete change in the way we look at data. By the way, this is not an IT problem only; it is an organizational problem and, even more, an IT industry problem. It may be time for vendors, user organizations and others to get together to solve this problem.
