When someone says "300 GB drive," you immediately think ATA, right? That may not be the case for long. Hitachi Global Storage Technologies, the disk drive manufacturing arm of the multinational giant, has announced a 300 GB enterprise-class drive, spinning at 10,000 rpms, with either Ultra320 SCSI or 2 Gbps Fibre Channel interfaces.
According to Doug Pickford, HGST director of market and product strategy, "the UltraStar 10K300 is perfect for traditional applications that continue to want higher and higher capacity points" (in other words, databases). Pickford predicts that "the drive will be a very needed and very important product in our portfolio."
For the past couple of years, the capacity heavyweight has been the ATA drive, and now the serial ATA drive. Manufacturers such as Maxtor Corp. are shipping drives as large as 320 GB, which are targeted at applications such as nearline storage and disk-to-disk (D2D) backup.
The UltraStar 10K300 breaks the capacity record for the SCSI drive, set over three years ago by Seagate, with its Barracuda 180. This 7,200 rpm, 181 GB drive was discontinued in 2002.
Hitachi achieved 300 GB capacity by more than doubling the areal density (the amount of data can be stored per square inch) of its platters, rather than increasing the number of platters per spindle.
Other manufacturers of high-end disk drives are expected to announce competitive products in the coming months. Fujitsu's vice president of marketing, Joel Hagberg, says the market can expect an announcement within the quarter. Additionally, Seagate will introduce a "fat Fibre" drive sometime this year, according to a company spokesman.
Does the UltraStar 10K300 set the stage for even bigger SCSI and Fibre Channel drives down the road? "It's hard to say at this point," Pickford says. "The idea of 600 GB in a single spindle exposes the user to the possibility of a double failure." If a drive is being rebuilt and a second drive fails, all of the data will be lost.
Network Appliance indirectly acknowledged this problem in December, when it announced RAID-DP, which stands for dual parity. Shipped as the default setting on NetApp's NearStore line of low-end filers, RAID-DP calculates two different types of parity (row and diagonal) and stores them on two separate dedicated parity drives. In that way, systems can survive drive failures (or hard errors) during RAID rebuilds.
Enterprise drives are not immune from this problem. "The hard error specification [the likelihood that a drive will produce an error] isn't any different for SCSI and Fibre Channel drives as it is for ATA drives," says Jerry Lopatin, NetApp vice president of storage systems engineering. These events occur every 1014 bits or 1015 bytes, regardless of platform. Assuming 500 GB drives, NetApp engineers calculate that, without dual parity, 5% of RAID rebuilds would fail.