Coping With SATA Growth

Bigger data volumes call for bigger disk drives, right? Well, storage managers are finding it's not always that simple. Sources say fatter drives can thin out performance. (See Capacity Considerations.)

To start at the top: There's a move among vendors to upgrade existing drives inside storage systems to faster, higher-capacity ones. Pillar Data Systems, for instance, has added 750-Gbyte SATA drives from Seagate to its Axiom Systems, which up to now have used 500-Gbyte Hitachi SATA drives. "We have increased the our overall capacity within a single system by 50 percent," Pillar CEO Mike Workman told Byte and Switch earlier this week. (See Pillar Pushes Provisioning, Capacity.)

And 750-Gbyte drives are just an example. With 1-Tbyte SATA drives from Hitachi and Seagate on the boil, the ante has been upped. Last week, Nexsan reported the adoption of Hitachi's 1-Tbyte drives in its storage arrays, upping capacity 33 percent. (See Nexsan Deploys Tbyte Drives and Idealstor to Add 1-Tbyte Disk.)

Some analysts warn that swapouts can stymie performance. "It's not that seek time, access time, or latencies change as drive capacity increases," says Arun Taneja of the Taneja Group consultancy. "All that is going to be the same. In fact, the native performance of a drive may even improve. But it's a fundamental truth that as you add more capacity behind a single drive, designing for system performance without increasing cost becomes more challenging."

"Higher capacity drives have the same number of heads as lower capacity drives," says one storage consultant, who asked not to be named. "So the number of heads per gigabyte declines, and if nothing is done to increase controller performance, performance declines.""Surprisingly, you may actually find some power savings from shifting from older, more power-hungry 500-Gbyte or even 250-Gbyte disk drives to newer, more energy-efficient 750-Gbyte SATA disk drives. On the other hand, when you move from, say, a 146-Gbyte disk to a 500-Gbyte or 750-Gbyte disk drive, then there are a lot more things to consider and keep in mind," states Greg Schulz of the StorageIO consultancy. "You want to keep in perspective how much more data you will be storing and subsequent rebuild times when a disk fails."

According to analyst Mike Karp of Enterprise Management Associates, the rebuilding of data in a RAID 5 environment can cause huge bottlenecks if capacity has been added without consideration for tweaking the controller design. "It's a huge issue," he says.

One user concurs. "Obviously, a 750GB drive is going to take a lot longer to rebuild than a 146GB drive. So, the potential window for a secondary disk failure in the same array is possible. Depending on how the storage system is architected and utilized, it may take as long as a couple of days for a rebuild of a 750GB drive," maintains Steven Olson, infrastructure manager at the Las Vegas Review-Journal newspaper. "That is a fairly large exposure. This is of course why RAID 6 has become so popular in SATA-based large capacity drive systems."

Vendors claim their designs compensate for potential performance glitches as SATA capacity grows. Pillar, for instance, offers what it calls "distributed RAID," in which a single RAID controller is shared among disk drive modules (called "bricks"), but each brick gets additional processors as capacity grows. "When we add capacity, we add more processing power," says Pillar spokesman Chris Drago.

Other vendors claiming SATA-friendly designs include a range of newer SAN and NAS players, such as iSCSI players EqualLogic, Intransa, and LeftHand Networks, as well as newer storage suppliers like BlueArc and Compellent.Most of these suppliers say they can ensure data can be spread across multiple drives at once, instead of being limited to a set number of drives.

"A supplier should spread the workload across a pool of drives, with no individual RAID sets," says John Joseph, VP of marketing at EqualLogic. He also advocates the use of more drives with lower capacity when throughput is a consideration.

Compellent's position is similar: "Modern SANs such as Compellent's will accommodate the larger drives with negligible impact on an end-user's overall experience by virtualizing the storage system - treating all the disks in the system as one virtual pool - and enabling automated tiered storage," writes Compellent VP of marketing Bruce Kornfeld in an email today.

One implementer acknowledges the importance of addressing all drives simultaneously. "In general you want to level the load, make sure all drives are spinning at once," says Miles Feinberg, founder and president of Miles Consulting Corp., a Calif.-based IT consulting and outsourcing firm. "Hard drives are the slowest things we have in the network." Ensuring that all disks share processing can help with newer RAID implementations and can avoid bottlenecks.

The ability to address multiple drives is also key because many vendors advocate adding a number of drives instead of making individual drives larger. If all that's needed is more capacity to store archived video, for instance, swapping a larger drive in may be just fine. But if throughput is required to handle transaction-oriented applications such as email, it makes more sense to simply add a larger number of drives."The basic rule of thumb for performance is larger quantities of smaller drives. This ultimately raises the spindle count, the number of heads, and the amount of aggregated device cache. This will lend to maximizing the available bus bandwidth," states Olson.

One user says the tradeoff takes consideration. "I think there's a lot of thinking that needs to go into 'Do I upgrade just the drives or the entire array?' I've struggled with that logic myself lately," writes Alan J. Hunt, manager of operations at the law firm of Dickinson Wright PLLC, who also happens to be an EqualLogic customer, in an email today. "[The] benefit I get from just replacing drives for larger ones comes down to whether or not I want to maintain an additional array or if I'm intentionally adding tiers."

Good design calls for more than clever technology, says StorageIO's Schulz. "You need to align the right disk drive to the right application and service requirements... striking a balance between performance, availability, capacity, and energy consumption."

Mary Jander, Site Editor, Byte and Switch