Disc technology is here to stay

Despite rumours to the contrary, disc storage will be with us for some time yet

Despite rumours to the contrary, disc storage will be with us for some time yet

For all the talk about alternative storage media, magnetic disc drive technology will remain the predominant online subsidiary storage medium for the foreseeable future.

In the opinion of Currie Munce, director of storage systems and technology for IBM's Research Division, this will mean traditional drives will be with us for at least the next seven years, although he believes it will prove to be much longer.

Until recently, it was thought that the IT industry was approaching a barrier that would prevent any further development of disc technology. That barrier is known as the "superparamagnetic limit". It is the point at which magnetic domains are so small they become unstable, rendering the data they represent liable to corruption.

Research carried out just four years ago put that limit at a packing density of between 20 and 40Gbits per square inch. Disc drives are now available using a packing density within those limits. IBM's Travelstar 30GN drive, launched in March, stores data at 25.7Gbits per square inch.

In the past four years technology research has made major steps forward. "Today, most people looking at magnetic recording are talking about a limit of 400Gbits per square inch, possibly up to 1Tbit per square inch," says Munce. He does not rule out the possibility that the superparamagnetic limit will prove to be 10 times as dense as that, although he says, "Some breakthrough invention or thinking will be necessary before then."

Packing densities on disc drives increased at 75% a year throughout the 1990s. At that rate, they would reach 400Gbits per square inch in five years, and Munce's top limit of 10Tbits per square inch in just 12 years.

Munce concedes that it is unlikely the industry will continue to develop the technology at the rate it did in the 1990s, pointing out that 75% growth rate is much faster than Moore's Law, which governs the rate of semiconductor technology progress.

Formulated by Intel co-founder Gordon Moore, Moore's Law states that the performance of integrated circuits doubles every 18 months - an annual growth rate of just over 60%. It has proved more than adequate to predict the breakneck pace we are all familiar with and power the computer industry forward.

So the disc industry can afford to slow its pace of development without seriously affecting users' ability to do what they want with IT. Indeed, throughout the 1980s the pace of disc drive development was much slower - just 17% a year. At that speed, it would take a decade to reach disc packing densities of 100Gbits per square inch - much less than we now believe to be possible.

The actual rate of development over the next decade is likely to be rather less than the industry managed in the 1990s, but more than in the 1980s. Munce says IBM is investigating a number of possibilities that have potential for further progress.

These include perpendicular recording, where the bits are oriented vertically rather than horizontally and thus can be packed more closely together; thermally assisted writing, which allows use of harder materials which can record smaller bits; and an improved recording layer with reduced noise levels, which requires a smaller number of grains of material for each bit stored.

Improvements in disc packing density have been matched by gains in price/performance, with disc subsystem prices falling from £9 per megabyte in 1990 to 15p per megabyte today. A major challenge for any technology looking to replace disc is to match first the price and then the rate of price improvement.

The technology which is considered by many to offer the greatest potential is holographic storage. This is because a hologram stores data in three dimensions - magnetic disc and most other storage technologies use only two dimensions. The potential for storage capacity is thus much greater.

For this reason companies have been trying to develop holographic storage devices for well over two decades. The US government has sponsored industry consortia research projects through its Defence Advanced Research Projects Agency (Darpa). Companies that have undertaken research include IBM, Rockwell and Bell Labs. Earlier this year Lucent formed a new venture, InPhase Technologies, specifically to develop holographic storage systems using technology developed at Bell Labs. A number of start-up companies have also thrown their hats into the ring.

The results of these efforts have been promising enough to keep investors interested. A Darpa project achieved a storage density of 250Gbits per square inch in 1999 - 25 times the level that contemporary discs are achieving.

Getting technology to work in the lab is one thing, producing a commercially viable product is quite another. So far, no practical and cost-effective device has emerged. Laser technology is well enough understood, but finding the right storage material has been difficult.

"The challenge has been finding the right recording medium with all the right attributes. Some materials have one or two, but not all the five, six or seven you need. That still remains the case today," says Munce.

"For a general-purpose replacement for rewritable hard disc, we need a new material to be invented. That could happen tomorrow." But, Munce thinks this is not very likely. "We have been looking for it for 20 years." he says.

Munce does not expect holographic devices to be competing with discs in the near future, although he believes they may emerge as special-purpose devices for niche applications.

That is what happened to two other technologies widely touted as disc replacements in the 1970s: charge-coupled devices and magnetic bubble memories. Both looked promising in the lab, but were never able to meet the fast-moving cost/ performance target set by disc, and both were eventually restricted to limited markets.

In the meantime, IBM and others will continue to develop disc technology and, even if the pace of development does get very much slower, the discs themselves can get larger, says Munce. "Today, a typical desktop drive has one platter, so we can always add other platters. It will be easier to do this than come up with a new technology," he explains.

Read more on Data centre hardware