Capacity Considerations

If anyone out there thinks bigger disks are a panacea, storage pros will quickly set them straight. While hard drive capacity's on the upswing, it's a challenge to manage it effectively.

There's no question users need more space. Data is expanding like the Blob, threatening to outgrow all kinds of storage systems. In some sectors, data growth of 30 Tbytes to 40 Tbytes annually is expected. (See IBM Scales Up for Healthcare.)

Drive vendors like Hitachi and Seagate are responding to demand with larger drives, and 1-Tbyte versions are ramping up. (See Vendors Tilt Over 1-Tbyte Drives.) Equipment vendors are following suit with disk array upgrades, particularly ones with SATA interfaces. (See Pillar Pushes Provisioning, Capacity and Nexsan Deploys Tbyte Drives.)

Problem is, capacity increases have a downside. Maybe two or three.

First, there's the risk of compounding data errors. "It's a huge issue," says analyst Mike Karp of Enterprise Management Associates. "Disks are getting bigger, but you're not improving the media, you're just making it bigger. If you have one bad spot per million disk sectors, that's multiplied."The problem is "really scary," he says, in RAID 5 environments, where media errors can affect applications -- a key reason for the existence of RAID 6. (See Adaptec, Intel Team on RAID.)

Another problem is price. Sure, SATA and SAS drives are aimed at demand for lower-cost drives. (See Emulex Locks Onto SAS.) But as Pillar's latest upgrade shows, greater drive capacity equals greater cost, even with SATA drives. Then again, the math may work: On paper, Pillar is increasing disk capacity in its Axiom system by 50 percent in moving to larger drives, while increasing the price per "brick," or drive increment, by just 15 percent.

The key problem is performance. "Larger capacity can actually reduce performance," says one industry analyst, who asked not to be named. "Higher capacity drives have the same number of heads as lower capacity drives. So the number of heads per gigabyte declines, and if nothing is done to increase controller performance, performance declines."

This analyst also points out that SATA drives typically have lower RPMs than Fibre Channel or SAS drives, which translates to fewer IOPS (input/output operations per second).

Another analyst thinks it comes down to knowing how to apply capacity effectively -- and that's no easy task."Sure, you can improve power consumption with larger-capacity disk drives, however, you also risk negatively impacting performance, unless you are dealing with static storage where performance is not an issue," asserts Greg Schulz of the StorageIO consultancy. It's important to focus on how higher capacity drives are configured or where they are used, he maintains. It wouldn't be advisable, for instance, to use a slower SATA disk drive instead of a Fibre Channel drive for a high-performance application, even if capacity increased.

According to Schulz, it takes homework to use capacity increases to best advantage. "Align the right disk drive to the right application and service requirements and look beyond the classic dollar-per-gigabyte metric," he counsels. Users need to strike a balance among performance, availability, capacity, and energy consumption (PACE), he says. RAID levels will need to be adjusted accordingly. "For example, if you are concerned about disk drives failing or taking a long time to rebuild, look at a different type of disk drive, or, look at a different RAID or controller configuration."

Bottom line? Capacity alone won't solve anyone's storage problems. How capacity fits into a broader scheme of data center issues determines its ultimate benefit.

Mary Jander, Site Editor, Byte and Switch