As far as storage goes, IT shops have been in the catbird seat for years. In fact, as prices on storage disk drives continue plummeting and capacity skyrockets, data center denizens have become almost oblivious to the technical marvels that hard disk drive makers produce. Consider this: according to "Alchemy and Rigorous Science: The Evolution of Hard Disk Drives," a recent report from Gartner Dataquest, per-platter capacities on hard disk drives have grown more than 100 percent annually for the past several years -- outstripping the growth rate of even Moore's Law. The report goes on to state that from 1997 to 2006, average capacity per drive will jump more than 10,000 percent -- from 2.7G bytes to at least 283G bytes -- while the factory price will decrease by 70 percent, from $208 to $69.
"If you look at the upward slope of areal density combined with the downward slope of cost per megabyte in storage, it essentially defines the revolution of storage technology," says John Monroe, vice president of Gartner Dataquest's storage group and author of the report. "It's impossible to overemphasize the influence and meaning of these two slopes."
In other words, IT workers have come to rely on the fact that they can expect to pay less and less for more and more storage every year. But can this phenomenon last? After all, disk drive makers can only drive so much out of the costs of manufacturing and distributing their products. There are signs -- such as IBM Corp.'s decision to sell its disk manufacturing to Hitachi Data Systems Corp. -- that consolidation could occur as the market becomes ever more commoditized, says Jim Damoulakis, the vice president of professional services at GlassHouse Technologies Inc., a storage consulting company based in Framingham, Mass. "Manufacturing efficiencies are key for maintaining profitability in this industry," he says.
More important, analysts point out that developers can push current disk drive technology only so far before hitting a limit; this phenomenon is known as the "superparamagnetic effect." Jon William Toigo, a storage industry author based in Dunedin, Fla., explains the issue this way. "The problem is the limitation on how closely you can pack bits together on a platter before the magnetic energy wears off," he says. "When the magnetic energy that holds bits in the recorded state becomes equal to the thermal energy generated by the drive itself, the bits flip from 1 to 0 and vice versa," he says. "This is the superparamagnetic effect. It's a good way to spoil your whole day."
The question is: how much more densely can companies pack those bits before bit flipping occurs? Toigo says that with current technology, disk makers should hit the limit of areal density at about 150 gigabits per square inch, compared with today's density of about 100 gigabits per square inch. "At the current rates of growth, we should hit that in the 2005-2006 time frame," he says. Monroe gives it a little more time, saying that he thinks that by 2006, companies will be able to pack nearly 300 gigabits per square inch onto each square inch of disks.
What other technologies under development might eventually supersede current disk drive technology? There are several interesting ideas, says Toigo, although he predicts that none will be ready for prime time before disk densities max out. He cites in particular.
- Heat-assisted magnetic recording (HAMR): "Here, you soften the storage media up with heat, record the data and then let it harden," says Toigo. "You can imprint the media with data that isn't magnetic at all, and it's erasable, so it can be overwritten."
- Perpendicular recording: This technology packs bits in more densely by sticking bits into the recording medium at right angles. "If you can point electromagnetic fields downward instead of pole to pole, you can pack bits closer together," says Toigo.
- Atomic force microscopy recording: Developed by IBM and code-named "millipede," this technology uses nanoscale tips to make small indentations in a plastic film, one per bit. The film can be re-recorded, and the device could eventually record at the individual atom level.
Promising as these technologies are, they still need time to develop mainstream appeal -- and price points. So, says Monroe, hard disk drives will remain the primary data center storage vehicle for the next decade. He does caution that the rate of development should begin to slow by 2003. Not to worry, however: the slowdown means that the rate of disk drive capacity growth means that disk drive capacity will meet, rather than exceed, Moore's Law. In other words, it will still be growing very fast, but not at the super fast rates we are used to. Meanwhile, prices will drop more slowly as well. The Gartner report predicts cost reductions of 14 percent in 2003, nine percent in 2004 and 2005, and a mere four percent in 2006.
They won't just be with us; they'll remain a key storage medium.
"I believe that hard disk drives will be our primary storage component for at least 10 more years. Nothing else can compare in terms of cost-capacity performance," Monroe says.
For more information:
- Check out this recent data center futures article: "Moore's law has a shelf life: Chip makers plan for a post-silicon world."
- Search390.com's Storage Management Best Web Links contain a lot of info on disk drives.
- For another look at disk densities, read "The data density dilemma."