uckily, the end leads to the Oz of computer storage: polymeric holograms, clusters of manganese molecules, and individual atoms of uranium. Oh my.
Head Case
The most expensive, and probably the most complex, part of a hard drive is the head that reads and writes data on the platter. Since heads are the primary cost of a hard drive, and since hard disk sales depend primarily on cost, the cheaper the head, the more successfully the drive will sell. This is not exactly a motivation for vendors to experiment with new, but pricier, technologies. Buyers could care less what kind of heads are in the drive: They are looking at the bottom line. (In this same issue, you can get the latest information and performance ratings of today's best hard disks. Our Lab Report tests will
help you pick the best drives for workstations and servers, in capacities ranging from 4 to 23 gigabytes. See the BYTE Hardware Lab Report "15 Disks Cover More Data Than Ever".)
Magnetoresistive (MR) heads continue to supplant older inductive heads. (Magnetoresistance is the property by which a material's resistance to electricity changes in a magnetic field.
Magnetoresistive
heads use this property to detect magnetic fields on hard disks.) Drives with magnetoresistive heads now claim some 50 to 60 percent of the market, and their share is growing, according to Jim Porter, president of Disk/Trend (Mountain View, CA), an analysis company that tracks the storage industry. Not surprising, given what is possible with this "older" technology. Already, MR drives are available from Seagate and IBM that hold more than 18 GB in a 3-1/2-inch form factor and more than 45 GB in a 5-1/4-inch form factor. By the time you read this article, other companies will probably also be offering MR drives
. Tweaking MR drives further (by diminishing the distance between MR head contacts, already measured in microns) will probably give them another 3 to 4 years of life. By then, MR will probably have run its course and not be able to support the higher
areal densities
possible with newer drive technologies.
What kinds of technologies? The most likely contender is the
spin valve head
, which uses giant magnetoresistive
(GMR)
effects. IBM developed GMR and spin valve heads, which rely on the exquisite sensitivity to magnetism of a thin conducting (but nonmagnetic) layer between conducting and magnetic layers. Spin valve heads increase the areal density some 10 to 20 times that of MR heads, to 10 to 20 gigabits per square inch.
Naturally, the GMR heads cost more than the MR heads, which is one reason they've been biding their time in the wings for a decade (IBM first discovered the effect in 1988). Still, it seems like their time has finally come. In
November 1997, IBM released its very first hard drive using GMR technology; it stores 16.8 GB in a 3-1/2-inch form factor for PC desktops (2.6 gigabits per square inch). IBM expects to move all its hard disk products to GMR technology. Analyst Porter thinks it likely that more GMR heads will make their commercial debut in 1998, probably from IBM, Fujitsu, Hitachi, and Seagate. These devices could initially have 10 to 20 times the capacity of MR drives.
As Porter points out, this is all part of the historical trend in areal densities of hard drives. Areal density has been growing about 60 percent per year, and that trend seems to be continuing. One implication is that areal density increases tenfold every five years. Within five years we could very likely have hard drives that hold hundreds of gigabytes.
GMR and spin valve technology are not the end of the line for hard disk drives. Already scientists at IBM are talking about "colossal" MR, and MIT researchers are at work on tunnel junction magne
toresistance. Both of these technologies could provide more storage density than GMR.
There is, of course, a limit to what you can record and read magnetically. This is called the
superparamagnetic limit
. Basically this means that the tinier the magnetic domains are in which you record information, the less stable they become. Eventually these domains would become so small that they would not be stable enough to hold information reliably. Engineers like to argue about what this limit is, but certainly something on the order of 100 gigabits per square inch is a reasonable high-end figure. Note that the hard disk technologies described earlier are rapidly approaching this limit.
Still, there is clearly more ore to be mined in the rotating magnetic hard disk. But there are also some other storage technologies that branch radically into different directions.
Packing Light
Holographic storage is one of those technologies that experts say is about five years from becoming practic
al -- no matter when you ask them. As of today, that would be after the turn of the century, around 2003.
This is not because holography is something new and mysterious. On the contrary, holography itself was discovered shortly after the invention of the laser. To form a hologram of an object, you shine one low-power laser at the object and an identical (reference) laser at photographic film. The laser light bouncing off the object forms an interference pattern with the reference laser light, and the photographic film records that interference pattern. The photographed pattern looks nothing like the object: It looks like gray and black smears and splotches. But when you shine a laser beam through the pattern (like a slide in a projector), you can see the entire object -- in three dimensions. Weirder still, you can cut the film in half and the remaining half could still reproduce the entire original image.
Eventually scientists figured out how to make the 3-D image visible to the eye with ordinary
light (a white light hologram). Now these 3-D holograms appear on software boxes and credit cards to signify that they're genuine.
None of this has anything to do with holographic storage, except for one thing. The idea that you can take three-dimensional images (data) and save them in a two-dimensional format (hologram) suggests an incredible saving of space to store information. It would be like replacing your file cabinets with pictures of them painted on the wall.
Turning this into a practical storage system means overcoming several obstacles. For one thing, lasers started out as large, delicate, expensive, and ungainly pieces of laboratory apparatus, while a viable storage system would need to take the rough-and-tumble life of computer equipment. Second, since you presumably want to be able to record and read data many times, you need a material that can become dark or light -- and stay that way until you change it -- over and over again. Third, this material would have to be accurate, reliab
le, and quickly accessible. Finally, it would have to be cheap.
Each of these obstacles is being knocked down by advancing technology. For example, every portable CD player has a laser in it, in the form of a solid-state laser diode. Since these laser diodes routinely withstand motions like jogging, dancing, and reckless driving, the relatively sedentary existence of a computer should not be a problem.
At first, the problem of a suitable medium seemed to depend upon naturally occurring crystals of photosensitive substances, which would be quite expensive to find and use for holographic storage. The problem of developing a suitable storage medium has brought the Photorefractive Information Storage Materials (PRISM) program into existence, paid for by the U.S. government (natch) and a group of interested companies including GTE, IBM, and Rockwell. Research has concentrated on difficult-to-produce artificially grown iron-doped lithium niobate, strontium barium niobate, bismuth silicate, and barium ti
tanate crystals.
It might also be possible to use special polymers as the photorefractive medium. Dr. William E. Moerner discovered photorefractive polymers while working at IBM's Almaden Research facility. In these special polymers, laser light causes an optical nonlinearity that allows the formation of dynamic holograms. The original hologram was 125 microns thick, and two laser beams could write and read the information. Dr. Moerner, now at the University of California at San Diego, is working toward improving this polymer-based holography system, including increasing the time that the polymers can retain the images.
These photorefractive polymers may change the whole holographic ballgame. For one thing, they will probably be cheaper than the previously used inorganic crystals. Also, polymers are easier to make and mold into different shapes.
How voluminous could holographic storage be? Eventually you could store the
Encyclopædia Britannica
on a hologram the size of a dime.
Theoretically, the potential storage density could be as high as 1 terabyte per cubic centimeter (about the size of a sugar cube). More practically, densities of 10 GB per cubic centimeter seem reasonable. A terabyte would fit in the space of a paperback book, or a few cubic inches.
Access would be fast, too. Since holograms store data in two-dimensional planes (or pages or sheets), reading a hologram delivers an entire plane of information. The throughput of such a system could be on the order of 1 Gbps. That's fast, much faster than hard drives of any type, and faster than most computers can handle. Time to bulk up that RAM cache.
One strange advantage of holographic storage has to do with the nature of holograms themselves. As noted earlier, you can lose half of a 2-D hologram and still recover the entire image. A similar effect is true with holographic storage. As opposed to hard drives, where losing a single stored bit can ruin your whole day, losing a dot in a hologram doesn't affect your
ability to access your data at all. You still get everything you stored.
Itsy Bitsy Magnets
If you want to use very tiny magnets, you can't get smaller than a few atoms. That is precisely what researchers have done with what are called
nanomagnets
. For example, a collaboration of scientists in Florence, Rio de Janeiro, and Grenoble has been working with clusters of manganese ions. In particular, at temperatures of 4°K, 12-ion clusters of manganese retain their magnetization for some time. Obviously, that would be important for storage of data.
What is even more interesting for storage, however, is that this ion cluster shows
hysteresis
. This means that applying a particular-strength magnetic field to the clusters causes them to take on one of two magnetic states. This binary-type behavior is precisely what computer storage requires. And the clusters of ions, on a near-molecular level, imply amazingly large densities of information.
Naturally, there are many tec
hnical obstacles to this type of storage. Since the clusters themselves are so small, it is difficult to manipulate them. What's more, there is no easy way to access single clusters, either to apply an external magnetic field selectively or to read the state of a cluster.
Let Me Atom
What could be smaller than clusters of atoms? Single atoms. The atoms would have to be powerfully ionized in order to have detectable effects. Fortunately, nature provides atoms that can become powerfully ionized. Unfortunately, the atoms are uranium.
Uranium has the highest atomic number of any naturally occurring atom. Since uranium has such a high atomic number, stripping away its electrons leaves it powerfully ionized. Scientists at Lawrence Livermore National Laboratory have been able to do exactly this. If the process could ever be made practical, these highly ionized atoms could form the basis for storage on an atomic level.
Clever people keep developing clever ideas for computer storage, from pr
oducts you can hold in your hand to concepts built around something as small as an atom. Whether all these great ideas ever turn into viable technologies is not important. What is important is that some of them definitely will. Lucky for the rest of us.
Where to Find
IBM
Armonk, NY
Phone: 800-426-3333
Phone: 914-765-1900
Internet: http://www.ibm.com
Storage Part 2: Infinite Space
