A few days ago I sat through an SNW session presented by the venerable Scottsman Willis Whittington of Seagate. He even had a “beam me up Scotty joke”. A great time was had by all.

Willis went through a good comparison of desktop drives and enterprise drives. I think a lot of these differences are well understood, but just for some quick background: ATA has historically been used in the desktop and SCSI/FC in the enterprise. Recently the ATA interface has migrated to SATA and SCSI has migrated to SAS. Desktop drives are big (capacity-wise), cheap and fail often. Enterprise drives are smaller, expensive and fail less often. That’s a gross over-simplification, but I think it works.

(BTW, those characteristics have nothing to do with the drive interface. They’re just arbitrary differences created by the drive industry. But enough about that…)

I’m not going to bore you with another discussion of the differences in desktop and enterprise drives. Instead, I will bore you with a discussion of a new type of drive called “nearline”. These nearline drives have actually been around for a year or two, but their actual definition has been debatable. I think everyone would agree that nearline drives have the low cost and high capacity of desktop drives but with the higher reliability of enterprise drives. They’re typically used for storing large amount of data that doesn’t have to be as fast as the local enterprise drives but they do need to be reliable and always online.

Well, Willis presented a very good overview of how Seagate’s nearline drives are different from their desktop and enterprise drives. Of course other vendors are free to choose their own definition of nearline drives. But I like Seagate’s definition.

The first difference that jumps out is the Mean Time Before Failure (MTBF), or the total lifetime of an average drive. Note that the MTBF of a drive comes from the manufacturer and the numbers have somewhat of a marketing slant to them, so take them with a small grain of salt.

Desktop drives are spec’ed to have an MTBF of 0.6MHours while enterprise drives are spec’ed to have 1.6MHours. Note that the desktop MTBF is not for a 24×7 duty cycle like enterprise. Desktop typically has a duty cycle of 8×5, or something similar. Nearline drives are somewhere in the middle at 1.2MHours.

The next difference is Rotational Vibration (RV) tolerance, or the tolerance of the drive to external vibration from other drives in the cabinet. RV is measured in radians/sec, but don’t bother trying to understand what that means. All that matters is that bigger numbers are better than smaller numbers.

So a desktop drive is spec’ed to run up to 6 rad/s, while an enterprise drive is spec’ed for up to 20 rad/s and often much higher. So what the heck does that mean?

Assume that a drive with no external vibration will run at 100% of its spec’ed performance. In the Seagate world of desktop drives that same drive will run at 80% performance if presented with an RV of 6 rad/s. It’s interesting to note that there is a quick drop-off beyond the rated 6 rad/s, with a desktop drive getting 10% performance at 20 rad/s.

In comparison, an enterprise drive gets 100% performance at 0 rad/s and 99% at 20 rad/s. There’s hardly any dropoff. In fact, the enterprise performance only drops to around 95% at 60 rad/s.

Of course the purpose of this post is to talk about nearline drives, and they are rated at 12.5 rad/s – again somewhere in between desktop and enterprise.

The next difference in desktop and enterprise is the amount of data protection (or error correction) in the datapath. Desktop drives have next to none, while enterprise drives have various types of protection including IOECD. IOEDC is used to protect the data in the drive buffer via parity or ECC. This enterprise feature has been incorporated into nearline drives.

Error recovery is probably the most well known feature of nearline drives. Typically a desktop drive is used in a solitary, non-RAID environment. Often this drive contains the only copies of your mother-in-law’s Orlando vacation photos, and you better not lose them! So a desktop drive will do whatever it takes to recover data. This can mean 30-60 seconds of retries. For a home user, that’s no big deal – just sit there and wait.

But if these drives are used behind RAID controllers, the controller will probably give up long before the drive – typically closer to the 10-15 seconds that are common on enterprise drives. Since nearline drives are often used in large disk farms that contain critical data it’s common to use them with RAID controllers. Therefore nearline drives have error recovery timeouts similar to enterprise drives.

Workload management is a new feature that is typically found only in nearline drives. These drives are often used in very densely packed disk farms where heat can be a major problem. So nearline drives will detect a specific temperature threshold and will enter a mode where every write command is turned into a read-after-write command. This simply means that after every host write the drive will read the data back to make sure it was written correctly.

This read-after-write does two things. First, it interjects a full disk rotation delay (4-8ms) into every host write. Since seeks are the major cause of heat, this delay significantly reduces temperature. A second benefit is that write data often becomes corrupted in high temp conditions, and the read-after-write will verify that the data was written correctly. If not, it’s re-written.

Next there’s the issue of power management. Desktop drives aren’t accessed as often, and therefore automatically spin down to save power and reduce noise. Enterprise drives have a much higher duty cycle and can’t typically be spun down, unless they’re spun down by command from the RAID controller or OS. Unexpected delays due to spin-up can cause undesirable latencies in accessing data.

So nearline drives do something in between. They power down certain circuitry but they don’t actually spin down the drive. This circuitry can be re-powered in almost unnoticeable timeframes.

Microcode download is important with enterprise drives. SCSI, SAS and FC drives have always had well-known commands for flashing new code to the drives. Enterprise drive users always expect the utmost in reliability, and unfortunately this is often achieved by upgrading the drive firmware.

On the other hand, expectations are low for desktop drives and therefore it’s very rare for anyone to upgrade the firmware. If a drive doesn’t work it’s easier to throw it away and buy a new one. These drives are practically free compared to the price of an enterprise drive.

Nearline drive are used in a manner more similar to enterprise drives, therefore they need the firmware download capability. And since there are often so many of these drives in the disk farm, it’s impractical to flash and reboot them one at a time. Therefore nearline drives typically support a feature where the new microcode image is staged to a reserved section of the disk, and is then read and flashed during the next restart. This allows all of the drives (dozens or hundreds, perhaps) to be flashed in unison.

Lastly, there’s the Write Same command. This is a command used by RAID controllers to more efficiently initialize an array – for example writing all zero’s to the drives. Again, nearline drives are often used in arrays so this is a useful command.

So, that’s the story of nearline – at least according to Seagate. But it’s a good story, and I buy it. Of course when you’re looking at other vendor’s nearline drives not all of these features may exist. But they’re probably pretty close.

Enjoy,
TT

This entry was posted on Thursday, November 2nd, 2006 and is filed under Storage Interconnects & RAID, Advisor - Tom Treadway. You can follow any responses to this entry through the RSS 2.0 feed. You can skip to the end and leave a response. Pinging is currently not allowed.