This article can also be found in the Premium Editorial Download "Storage magazine: Who owns storage in your organization?."
Download it now to read this article plus other related content.
About 300 years ago, Control Data's 205MB washing machine-sized disk drive sold like hot cakes to every major OEM (Digital Equipment Corp. called it the RA60). It cost more than plutonium (still available online for about a grand), but it had a removable disk pack (RMO5 in DEC-speak). All the disk platters were locked up in this cool Tupperware container that could be removed from the drive itself and replaced with a new pack.
Removable media meant disaster recovery! Content distribution! A load device! Sure, but which removable media? Clearly tape won. That's because disks blow up when you shake them, drop them, leave them in the sun, etc. And they were a million times more expensive than tapes.
The 3.5-inch drive revived interest in removable disk technology in the late 1980s. It was small enough and almost cheap enough. Capacity and performance outstripped tape. So users were willing to try shuttling them between primary and disaster recovery sites. At the end of the day, the same reliability problem of "oops, I banged the disk" outweighed the performance and portability gains of the removable disk.
But my little brain immediately started thinking of the possibilities opened up by removable drives. The RXT pack is about 4 by 2 by 12 inches, containing 12 2.5-inch drives, totaling about 1TB in capacity. The RXT is actually lighter than the same size tape cartridge. The smash-proof drives are shock-mounted in a canister. To boot, the drive unit that the RXT pack goes into reads and writes RAID onto the canister itself--either one or five. If and when a drive does go pop, the data can still be extracted.
If I can ship 1TB of fairly high-performance disk and have it actually work when it arrives at the other end, at a cost of around $6,000 per terabyte, I've just enabled disaster recovery for the masses! I just solved my content distribution issues! I'm a hero!
Let me reemphasize that this is me talking. SpectraLogic is not making such claims; it is sticking to its integrated VTL story. However, we all know that I'm not the only forward thinker out there, and some of you are going to buy this thing and do exactly what I've suggested. And when you do, let the rest of us know how it went, because eventually I'm always right--it's just a question of now, or another ten years from now....
This was first published in May 2004