The answer, and the pro-IDE arguments, are very straightforward: it's the economy, stupid. If you look at the ratio of available hard-drive capacity to price, SCSI drives make a pretty poor showing. The cost factor gets even more painful the higher you go. For the price of one 147 GB SCSI hard drive, you can buy three modern IDE hard drives, each with 200 GB and up. Now that's straightforward.

Note that we're talking about backups here. SCSI remains the undisputed champ of file-server systems where data availability is of utmost importance, since manufacturers have been tweaking these drives and controllers to offer the best I/O rates possible. The same cannot be said for the current batch of IDE drives, however. On the other hand, maximum performance is rarely needed in backup solutions.

Let's not forget our comparison with the classic tape drive, either. Robotic systems have no difficulty with vast quantities of data, but start-up costs for such a system are equally vast. Backup systems with robots will run a cool $50,000, or even more, and then you have the several streaming drives, the dozens or even hundreds of tapes, a backup server and the software to go with it. In short, this colossal investment will only pay off if you have so much data that you only have to pay for the tapes, which are relatively cheap.

Considered in that light, the reduced reliability of IDE hard drives is really not all that important. Any "true" RAID array also has a hard drive not being actively used. This "hot spare" kicks in the moment an active hard drive in the array fails due to some kind of defect. The controller recognizes the failure and automatically starts restoring the data on the failed hard drive using the hot spare. Again, let's look at the bottom-line aspect: according to our above calculations, you can buy either a SCSI hot spare with 147 GB or three 200 GB drives with IDE interfaces. And if one active drive were to fail, you'd still have two additional ones to fall back on.