% fortune -ae paul murphy

Disk drive price change

In June of 1992 I paid $3,100 Canabucks for a 330MB SCSI disk for a Sun workstation. In April of 1999 I paid $2,424 for a 9GB UltraSCSI disk for an HP C360. Right now I'm looking at $389 for each of two 143Gb disks for another Sun workstation. Ignoring performance issues, the price per megabyte in 1999 was about 2.8% of the 1992 price per megabyte and today's price is about 0.02% of what it was in 1992 - with better than two thirds of that price change happening in the second seven year interval.

Disk performance, mainly in terms of reliability and transfer speeds have gone up commensurately. That 3400 RPM, 330MB, disk could be expected to fail after fewer than 1% of the r/w operations its modern successors promise, and head settlement took longer on it than entire transactions do now.

By way of comparison low end SPARC stations have been in the $5500 range for the entire period with high end machines like like my dual CPU SPARC20 then or a new 2500 now running in the low twenties. Similarly Intel's CPU performance has gone up, but pricing has remaining surprisingly stable. Basically disk storage change has illustrated some exponential version of Moore's law, where CPU change has been merely marginally superlinear.

And it isn't remotely over. Last week Bob Cringely used his PBS pulpit to talk about a new disk drive technology he and some colleagues have been working on.

Here's his summary:

Two old friends of mine, Anil Nigam and Jim White, and their company, Antek Peripherals, Inc., had been working for years on technology for a sort of hyper-floppy drive using metal foil for the recording medium. At the time they were aiming their work toward digital cameras, but I asked if the same technology could be used in a non-removable form inside a computer disk drive? And if it could be used that way, what would be the effect on price, performance, reliability, and energy consumption? The answers were stunning: they could design new families of disk drives that held up to three times as much data in the same space, were more reliable, actually cheaper to build, and used 70-95 percent less energy to run than the current state of the art.

For a data center operator, simply swapping disk drives could increase storage by three times while decreasing total energy consumption by a third. This is a huge attraction for big businesses and big hosting companies, not just because it lowers their electric bills but because it may mean they can avoid building more data centers entirely.


This power savings is key and opens up whole new product categories, like tape replacement. Computer tape drives are a $10 billion business, but the dirty little secret is nobody really knows if they have the data or not, since tapes are not a reliable archival solution due to print-through and environmental deterioration. Disk drives would be better but disk drives cost too much. Not anymore. Our projected tape drive alternative costs $0.20 per gigabyte to tape's $0.18, but ours has greater reliability and a seek rate that is 720 times faster. The tape drive uses 18 watts while our tape replacement drive uses three watts.

Is his product real? I don't know -but the liklihood is that there's something here: enough to bring the cost per gigabyte down by another factor of ten or more - enough to make energy costs the dominant determinant of data storage costs.

Paul Murphy wrote and published The Unix Guide to Defenestration. Murphy is a 25-year veteran of the I.T. consulting industry, specializing in Unix and Unix-related management issues.