• United States

SATA devices now with vastly improved MTTF numbers

Apr 20, 20064 mins
Data Center

* SATA are robust enough for the enterprise IT floor

America’s great modern-day philosopher Yogi Berra once offered the following guidance: “When you come to a fork in the road, take it.”

Berra was of course an insightful man, a man who grew beyond some shameful early associations (he played for the New York Yankees) to develop a shrewd awareness that can often be usefully applied to managing your storage.

Consider the above quote, for example, placed within the context of disk drives.

Commoditized disk drives – ATA drives that have achieved commoditization because they have been sold by the tens of millions for desktop use – have reached a fork in the road. ATA drives have been around for at least 15 years, and those of you who have nothing better to do than to track the lifecycle of storage devices will certainly have noticed that the new generation of ATA drives – now with a serial interface (and thus referred to as SATA – “serial” ATA) last far longer than did their desktop counterparts.

But still, many managers express concern that SATA is still not robust enough to be put out on the floor of an enterprise IT room. Let’s put that misjudgment to bed right now.

These days the vendors are now producing SATA devices with vastly improved Mean Time to Failure (MTTF) numbers – numbers that now rival, and in at least one case, equal those we expect to see only on higher-priced devices that typically attach via SCSI or Fibre Channel.

Increased reliability has been accomplished through use of better parts – mostly, improved bearings so that the media spins with less friction, heat and vibration – but also through the addition of Native Command Queuing (NCQ). NCQ is a process through which a drive re-sequences outstanding I/O commands to put the commands in an order that reduces mechanical overhead. Because of NCQ, SATA drives now reorder to take into account the position of the drive’s heads relative to the data on the disk. This reduces head travel, which reduces physical stress on the whole drive.

Lessened head travel, heat, friction and vibration all result in a MTTF for the better SATA drives that is now in the range of 1 million to 1.2 million hours, which is exactly where we find the MTTF for SCSI and Fibre Channel drives.

In the past, MTTF numbers were difficult to compare across interfaces because they were based on differing duty cycles (a disk drive’s duty cycle is the percentage of time the drive spends servicing I/O requests – in other words, the time the device actually does some work). Because enterprise-class drives have an 80% duty cycle at 24 hours per day, and drives for the desktop have historically been built to operate at a 10% duty cycle for 12 hours per day (drives on the desktop rarely have to work 24 hours a day) MTTF comparisons didn’t make much sense.

The new drives however, are measured against duty cycles ranging from 20% to 100%, and will certainly be quite appropriate for inclusion in devices that service streaming applications or other applications where sequential read or write activities will be the norm. For these, SATA devices offer a terrific balance between reliability, speed (we’ll talk about this some other time) and capacity.

SATA drives have reached a fork in the road. Some remain suitable only for use with desktop applications, while others have been beefed up to provide a fully reliable part of many enterprise applications. Just make sure that when you read those MTTF numbers, you understand the underlying duty cycle assumptions. Then, see if that duty cycle matches your application’s requirements.

*** The 2006 edition of The Great Storage Haiku Contest is now officially underway, and the first entry – not likely to be a winner, JN – has already arrived. See last Thursday’s column for the rules. Send entries directly to me. I’ll announce the winners in May.