According to our reader surveys, you’re very interested in trends such as disk-to-disk backup/restore, tiered storage, fixed-content storage, and other nearline applications. What’s making these trends possible, in part, are inexpensive disk arrays based on Serial ATA (SATA) disk drives. But as you start to peel back the onion you get into the old interface debates.
To illustrate this, talk to disk drive manufacturers about which drive interface is best for your applications. Better yet, don’t. For business, rather than technology, reasons the drive manufacturers are a biased lot. Each has strengths (and weaknesses) in particular segments of the market, and market share to either protect or get.
The argument often boils down to whether SATA is “enterprise class” or not. A drive manufacturer with a vested interest in either Fibre Channel or SCSI will argue that SATA, like parallel ATA (PATA), is essentially a desktop-class drive that’s not suitable for enterprise-class applications. (“Enterprise class” is never defined, but it implies high duty cycles in server- or network-based configurations. But what’s high duty cycle: 7×24, 7×12, 5×10?)
Meanwhile, a drive (or controller) manufacturer that is staking its future on SATA will argue that SATA is in fact enterprise class, sometimes going so far as to claim that SATA is on par with Fibre Channel and high-end SCSI in terms of performance and reliability. I won’t even get into the performance debate, but the reliability claims are typically “backed up” by figures such as Mean Time To Failure. For example, a lot of drive makers claim an MTTF of 1.2 million hours. Although these numbers are based on extensive lab and field testing, they’re still fairly meaningless.
And besides, failure rates have little to do with the interface; they have more to do with the quality of the drives and components. Similarly, performance has as much (or more) to do with factors such as the disk rotation rate (7,200rpm, 10,000rpm, 15,000rpm) than with the interface itself. Not to mention the size of the blocks.
As you get into the interface debates, don’t forget about FATA—a sort of Fibre Channel and ATA combo that purportedly combines the best of both interfaces. FATA would be merely an interesting footnote except that it’s backed by the largest vendor of disk arrays—Hewlett-Packard (which is also a big fan of Serial-Attached SCSI, or SAS, which I’ve left out of this discussion because SAS arrays won’t be available until at least late this year).
Although the interface debates are fun both from a technology and business standpoint, I think the bottom line is this: SATA is cheap (as much as 50% less than Fibre Channel or high-end SCSI) and will be fine for most applications short of 24×7 operation with high I/O rates and/or OLTP-like small block sizes. In other words, you should definitely consider SATA for applications such as disk-based backup, tiered storage, fixed-content storage, and other nearline applications.