The final stage, or Doomsday Recovery, is where we spin to tape. The way we work today doing incrementals during the week and full backups on weekends, means that over time we will have tons of copies of the exact same data, on thousands of tapes. This is not only a waste, it makes recovery even harder. Instead, we're going to re-think what it is we are trying to accomplish. Since 99.7%+ of all recovery actions will occur on the previous two stages, this is really only for an outrageously small percentage where we would need to use this tier for recovery. Therefore, it makes no sense to have more than say, four copies of any unique data object. Once we have four copies on four tapes, we flag the data so that the backup system no longer makes any more backups of that data. In essence, we archive the data and remove it from the backup process. Doing this will most likely result in a 90%+ savings in media cost alone, which most likely would justify buying a whole lot of cheaper/slower disk based systems to keep all your real recovery data on.

The economics and the technologies required to do this exist now. This isn't fantasy. The most difficult element to come to grips with is mental - you don't have to keep doing things the wrong way just because that's the way we've always done them.

I'll get a full paper out on this soon, with Heidi Biggar's help, and we'll put a model in it that you can fill out to see the economic, performance, and overall impact a change would mean, by different products and what each different technology will mean to your operation. Soon. I promise.

Send me your questions -- about anything, really, to sinceuasked@computerworld.com.

Steve Duplessie founded Enterprise Strategy Group Inc. in 1999 and has become one of the most recognized voices in the IT world. He is a regularly featured speaker at shows such as Storage Networking World, where he takes on what's good, bad -- and more importantly -- what's next. For more of Steve's insights, read his blogs at thtp://esgblogs.typepad.com/steves_it_rants/.