Seagate is betting millions that Joe Davis can reengineer magnetic drives to sidestep the impending storage barrier.
Ten years ago you paid US$10 for a megabyte of magnetic storage. Today you pay about 10 cents. If you buy a computer next year, the hard drive in it will have twice the space for the same price. Like Moore's Law, you take it for granted, right?
Wrong. When it comes to storage, that assumption is about to collide with a law of physics. Storage capacity is related to the number of bits crammed onto your disk: Shrink the bit size and you have room for more data. But in a few years, the bits are going to get so small they'll no longer be stable at room temperature - the polarity, which designates whether a bit is a 1 or a 0, will flip spontaneously from north to south. Your data, in other words, will begin to self-erase.
"Don't believe companies that say this is not an issue," warns Joe Davis, cofounder of Quinta and a 25-year veteran of the disk-drive industry, with stints at Applied Information Memories, Maxtor, and Iota Memories. "That's what you'd say, too, if you were a magnetic storage company."
Davis's brainstorm: Take the head used in optical drives, shrink it to one-thousandth its current size, and plug it into a traditional magnetic drive. Voilà! No more binary flicker. This works because an optical sensor uses a laser, and the focused heat of a laser locks an individual bit into a stable condition.
The technology is so compelling that last year disk-drive giant Seagate bought Quinta for $230 million. The Silicon Valley start-up will get another $95 million windfall if it delivers a working product by an undisclosed deadline.
"Without a doubt, conventional magnetic recording is going to run out of gas in the next five years or so," says Mike Casey, a Gartner Group research director who follows the magnetic and optical storage industries. "When that happens, what Davis is doing will begin to look very smart indeed."
DATA TRANSMISSION
Everyone from shipwreck hunters to oil-rig inspectors relies on underwater cameras and remote sensors. But transmitting pictures and data from the deep is tricky - cables are unwieldy and can snag, while lasers can't penetrate the murky depths. "Even the chirplike sounds that dolphins use are a rather inefficient use of bandwidth," says Oliver Hinton, a professor of electrical engineering at England's Newcastle University and a specialist in acoustic communications.
Hinton has solved the subsea data dilemma with a device that combines multiple ears and sophisticated signal processing to reliably send and receive wireless data at 16 Kbps. While taking a cue from dolphins, Hinton one-ups Flipper with an array of seven hydrophones to better snag incoming sound waves. He also built in phase-key shifting - the same technology telcos plan to use for fast transmissions over phone lines - to pack sound waves with data. The prototype is bulky, about the size of two PCs without the hydrophones, but Hinton believes he can cut that in half for a portable model.
Numerous engineering, defense, and oil companies, including Rockwater and Shell Oil, have bankrolled the project; a commercial product, says Hinton, should be ready by next summer. For an encore, the Newcastle gang is looking at acoustically linked sea-floor WANs.
DISPLAY TECHNOLOGY
Remember when Mom told you not to sit too close to the TV because it would ruin your eyes? Times have changed. With new dime-sized microdisplays, you actually position your eye centimeters from the screen. These high-resolution displays could soon replace traditional LCDs. No more scrolling through messages on your pager or PalmPilot. When a fax or email arrives, just peek through a magnifier mounted over the tiny screen and it's like looking at a high-res color monitor.
Several companies are marketing such screens, including MicroDisplay, which uses conventional CMOS processes to manufacture silicon panels coated with liquid crystal. While MicroDisplay hasn't inked any deals, it expects the first consumer apps to be a new generation of viewfinders for digital cameras and camcorders.
Meanwhile, a company called Kopin is transferring the pattern of their silicon panels to a glass substrate. The benefit is that a display system using glass requires less-complicated optics in its viewer. The Taunton, Massachusetts-based firm has joined with Fujifilm Microdevices to develop a display and OS for digital cameras and with Motorola to provide screens for its cell phones (see Kopin prototypes, right). Stanford Resources forecasts steady growth for micro-displays - it expects worldwide shipment of 7 million screens in 1998 to increase to 17 million units generating revenues of about US$400 million in 2004.
PORTABLE COMPUTING
Assemble a small team of top researchers from Digital Equipment Corporation, ask them to build a palm-sized computer with the power of a PC, and what do you get? Itsy. Carl Waldspurger of Digital's Systems Research Center in Palo Alto, California, describes Itsy as a cheap, open-platform handheld machine with enough power to do it all, even voice recognition. Itsy uses a stock StrongARM 1100 chip operating at speeds of up to 200 MHz, has room for 64 megs of memory, and runs on three AAA batteries. The Digital team has produced almost 100 Itsies so far and claims it will seed several university research groups with the device. Compaq, which took over Digital in June, has said little about Itsy's commercial potential. Note to Compaq: Good things come in small packages.
The history of computer science is the history of less than a dozen powerful ideas. But there's a new one, called flat transactions, that will define our immediate future and reverberate through everything we do.
The flat transaction is a simple way for a computer to represent things like reserving a flight or withdrawing money from a bank account. Today, flat transactions are the backbone of airline and credit card networks. But outside of a few highly trafficked systems, the notion of flat transactions has remained obscure. Now, thanks to the Net, they are poised to become as ubiquitous as bits and mips.
"It is impossible to build large distributed systems without using something like flat transactions," explains Jim Gray, one of the three or four people who could reasonably be said to have invented the concept. Gray is now a researcher at Microsoft, where programmers have incorporated flat transactions into the Windows NT 5.0 operating system, underscoring the implacable rise of the flat transaction model.
Everyday analog transactions can be amazingly complex and multidimensional. Just think about the baroque protocols that surround a drug deal. You don't want to hand over your money until you have the goods in hand. And the dealer isn't about to give you anything until he or she has the dough.
Similar problems arise when designing banking or airline reservation systems: Events must occur as if simultaneous, even if they are not. When you transfer funds, the withdrawal and deposit should appear as one action. Otherwise, the computer might do the withdrawal, then crash, leaving you broke.
To avoid such mishaps, these transactions must be "flattened." This is done by constraining them to several explicitly defined properties, including two that are critical for building large-scale distributed applications: atomicity and isolation.
Atomicity means that a transaction cannot be half-done. Most transactions, like atoms, consist of several parts, so keeping them whole requires some programming trickery. A flat transaction system must put an envelope around the substeps, then ensure that each is executed. If the substeps are not completed, it undoes them and calls the transaction incomplete.
Isolation means that each transaction must remain unique. There is no way, in other words, for two airline passengers to be assigned seat 3B at the same time. Since in many systems thousands of transactions must be processed at the same time, some computational gymnastics are required to keep all of them isolated. Special locks are put on data that is shared to ensure that only one transaction can read or modify that data at any given moment. Together, isolation and atomicity guarantee that millions of delicate transactions work smoothly. But "flatness" also limits what we can do with transactions.
"Let's say you're traveling to Cancùn," says Michael Stonebraker, CTO of database vendor Informix. "You'll be booking flights, hotels, and tee-times over several days and weeks. With flat transactions, all that is either one transaction - in which case every time you cancel one piece you have to cancel the entire thing - or else it's many, in which case if you get sick and decide not to go you have to undo each part independently." You can work around these problems, but that means adding the complexity that flat transactions were invented to alleviate.
It's doubtful an elegant transaction model for the Net will show up soon. Instead, we'll see slight improvements - Microsoft and Oracle have dedicated significant resources to this task.
The most important consequence of flat transactions may be in the way we adapt to their peculiarities. After all, we already live within a web of flattened transactions: Visa handles about 80 million a day. And on the Web an increasing number of the 20 billion hits per day are supported by flat transactions. No wonder then that some of the ideas behind this esoteric technology have seeped into popular thought. Undo, for example, stems from transaction processing, where every transaction must have an antitransaction. If you can undo an incomplete transaction, or a mistaken stroke, then why not other events - a posting on a Web site, or something you just said and wish you hadn't? Flat transactions are becoming too ubiquitous, and the underlying idea is too beguilingly reductionist, for them not to color how we think.
GENETICS
Genetics, says Stanford molecular biologist Joe DeRisi, is concerned not with decoding and cataloging the blueprint of any given cell so much as discovering the role of each gene in an organism's life cycle.
Today, this is done with the help of a microarrayer - a robotic device that can put up to 13,000 genes on a 1- by 3-inch chip. Stanford University Medical School's Brown Lab designed and built the first microarrayer prototype in 1995, and the school later licensed the technology to Synteni. But rather than sell the machine itself, Synteni found it more profitable to make the finished arrays and charge companies about US$5,000 per chip for research results. With perhaps a hundred chips needed for a single project, this put a critical new tool out of reach for those making the biggest strides in genomics - academic and institutional scientists.
So with Brown's permission, DeRisi posted the plans for building the microarrayer on the lab's Web page (cmgm.stanford.edu/pbrown/mguide/). Provided your research is strictly nonprofit, you can download the specs and then build the machine yourself. Construction will cost you 25 grand, but that's much cheaper than buying the arrays one by one from Synteni.
Tito Serafini of UC Berkeley's Department of Molecular and Cell Biology is constructing two of the machines. "The people at Brown did a very good job of making the plans user-friendly," says Daniel Emerling, a Berkeley post-doc student helping build the arrayers. "Anyone who has ever put together a kit from RadioShack could probably build one of these in their basement."
CIRCUIT DESIGN
Embedded in that snappy display on your laptop, PDA, or cell phone are thousands of high-intensity pixels, paired with thousands of tiny transistors. It's a fairly typical design but also extremely expensive to produce. That's one reason you haven't seen displays show up in everyday goods like soda cans, business cards, and clothing.
"But if the cost of light-emitting elements could be decreased," says Henning Sirringhaus, a physicist at the University of Cambridge, "it would open up a wide range of new display applications." Sirringhaus and fellow researchers at both Cambridge and Bell Labs believe they can greatly cut the cost of displays with a new polymer called hexylthiophene.
Essentially a plastic, hexylthiophene has been engineered to reach speeds close to the amorphous silicon currently used in its place. As a result, transistors made from hexylthiophene put out enough current to drive a plastic LED. These LEDs could replace many of the semiconductors, phosphors, and liquid crystals used in today's displays. Sirringhaus foresees small, flexible plastic displays for mobile phones, toys, and even newspapers.
BIOSCIENCE
Bruce Bryan, CEO of Prolume, recently unveiled a bright idea that enables everything from soft drinks to malignant tumors to glow in the dark. Since February, Prolume has been cloning millions of gene proteins from luminous jelly fish, squid, and other sea creatures in search of their light source. The tedious process has turned up five proteins that shine when an organic molecule is added for fuel. Prolume has produced a slimy BioToy called Alien Crystals, which glow when water and the special NanoFuel are added. According to Bryan, the plan is to funnel toy profits toward research for medical apps, including a light-emitting protein that can illuminate cancer cells during surgery. "What we have here," says Bryan, "is the molecular lightbulb of the next century."
Eighty-five percent of Americans with telephones have access to ISDN lines (International Data Corporation/Link) ... By 2002, 10 percent of international telephony traffic in Europe and the US will travel over IP (Datamonitor) ... The number of Internet appliances, such as Web-enabled telephones and PDAs, will increase by a factor of nine by 2001 to 43 million (IDC) ... Within one year, 50 million people will subscribe to wireless data service, with 400 million by 2004 (Matthew Desch, president of wireless for Northern Telecom; Bo Hedfors, president and CEO of Ericsson, US) ... In DVD's first year on the market, Americans purchased 437,000 players, almost twice the number of CD players purchased when they first became available in 1983 (Consumer Electronics Manufacturers Association)
CRUCIAL TECH
Hard Driver
Catching the Sound Wave
The Small Screen Gets Smaller
Matchbox Pentium
The World Is Flat
Grassroots Biotech
The Light Stuff
Beyond the Pale
Raw Data
