Feb 1, 1998 12:00 PM

Yet Another Gigabit Operation

By Steve G. Steinberg

| CRUCIAL TECH

| Yet Another Gigabit Operation

Thin Client? Fat Chance

DNA Tweezers

<h4>#### latest hot box in the network router wars will open the pipes for videoconferencing and immersive multimedia.</h4 Athis, and spam blocking, too.

Norking has become as fickle as fashion. Every season – demarcated by the spring and fall Networked+Interop conferences – a new start-up is anointed "the one to watch." The chosen one usually has just a prototype of its "hot box" (some kind of switch or router that can direct data to its destination at high speeds) and a slew of new marketing terms to rattle off.

Lyear's buzz was about Ipsilon and its flow-labeling technology. This season the gossip is about Sunnyvale-based Yago Systems Inc. – and about Ipsilon struggling to meet payroll, then being sold for a paltry sum.

BYago appears to be made of humbler, more durable stuff than the standard hot Valley start-up. From the self-effacing name (Yago originally stood for "yet another gigabit operation") to the low-rent – some would say decrepit – offices, Yago is, in the words of CEO Piyush Patel, "a down-to-earth engineering company." What Yago employees are trying to do, however, isn't.

Tpast two years have seen a major shift in how network routers are designed: more of the intelligence has been put into hardware instead of software. This kind of hardwiring allows for faster speeds, but at the expense of flexibility. Now Yago claims to have achieved both with a custom chip of its own design.

Arouters look at the header of a packet, determine its ultimate destination, and send it on its way. Because this has to be done very quickly that's about all most routers are able to do. Yago's box, however, is able to do much more in the split fraction of a second it has to examine a packet. For example, it might see a packet that contains email data from a known spammer and block it.

Hdoes Yago do it? "We didn't hire networking engineers, we hired microprocessor designers," says Andrew Feldman, Yago's director of business development. "For these guys, achieving gigabit speeds was the easiest thing they've ever had to do."

Y's first products will ship early next year.

<arc Spiegler</em

<a hAnother Gigabit Operation</a>S

ming in Data?

<Ls</a>

<Client? Fat Chance</a>

<Tweezers</a>

Generng reams of numbers is easy; understanding them is more difficult. But New Mexico's Sandia National Laboratories is honing a system, EIGEN/VR, to transform complex data into virtual, fully navigable environments. Highlights include data sets involving a brain tumor and a comet hitting Earth. A more practical application? "You can go inside a tire tread and see where it's overheating," explains Sandia's Peter Lamphere.

BPescovitz</p

<ref=her Gigabit Operation</a> r

nith Sonar

you'rerobot, being blind as a bat may not be such a bad thing. Rodolph (short for robotic dolphin) is equipped with three Polaroid electrostatic transducers that emit sound waves and listen as they echo off an object. The new system, designed by Professor Roman Kuc at Yale University's Intelligent Sensors Laboratory, is sensitive enough to determine whether a coin is heads or tails – and may be an improvement of artificial eyes. "Vis

puces a two-dimensional image of a three-dimensional object," Kuc explains. For example, a camera used to identify an authorized user of an ATM could conceivably be tricked by holding up a photograph. But sound waves wrap around an object, giving off a distinct pattern of echoes that the sonar sensor can be trained to recognize, such as barely noticeable facial gestures, enabling quadriplegics to interact with a computer simply by moving their mouths or wiggling their noses. "I t

s technology will be very useful in the sensor-fusion approach to solving the vision problem," Kuc says.

aanwyn <p

oCRUC</strong> </

r"httGigabit Operation</a> <a h"

ming <a h

wn humorgans. It sounds like something from TNT's MonsterVision, but it's now a commercial reality. In April 1997, Apligraf became the first manufactured human organ approved, in Canada, for commercial sale. Produced by Organogenesis of Canton, Massachusetts, the human-skin equivalent is currently awaiting FDA approval in the US. Seeded b

lfrom infant foreskins, Apligraf is grown in what have been dubbed skin factories. "The cells from a single foreskin can produce 200,000 units of manufactured skin," says Carol Hausner, a spokesperson for Organogenesis. That's enough skin to cover about 250 people. The process begins by introducing human foreskin cells to a solution of bovine collagen. In about 30 days, with minimal intervention, the cells organize themselves and form a two-layer upper and lower dermis, just like our own skin. "We provide the supportive environment; they do what they're designed to do," says Hausner. Applicat

skin equivalent include treatment of such common clinical conditions as chronic wounds from ulcers, severe burns, and skin surgery. Manufactured skin is on the forefront of a revolution brewing in medicine called tissue engineering. Other forms of engineered parts bubbling away in medical labs include connective tissue, blood vessels, organic liver-assist devices, and heart valves. "People think I'm joking when I say we grow human skin, until I whip out a dish of it from my briefcase," chuckles Hausner. "Skin's a fun tissue to work with." By S

nberg (<a href="http://org</a>) is a Wiredtributing editor and a consultant with a New York investment firm. |

CIAL rong> <h

"p://bit Operation</a> <a href=p

phref=

s are #### g to make wires the buggy whips of the next millennium. By Steve G. Steinberg</h4> For the pas ys, technology has been driven by steady improvements in lasers and silicon chips. Smaller and faster chips brought the power of a Cray supercomputer to videogame consoles; cheaper and more accurate lasers brought the CD player and high-speed fiber-optic networks. Now these tw

ns have been grafted together in what is probably the most important innovation of the decade: vertical cavity surface emitting lasers, or VCSELs (pronounced "VIX-els"). These tiny lasers – each less than half the width of a human hair – are fabricated the same way that computer chips are. They bring all the advantages of optics, such as increased bandwidth and reduced power consumption, down to the chip level. Just as we'v

eelephone companies rip out old copper lines and replace them with fiber-optic cables that can carry hundreds of thousands of times more traffic, now we're going to see the copper wires in our computers and electronic devices replaced by VCSEL-equipped fibers. Instead of dozens of wires connecting a graphics card to the motherboard, for example, a couple of tiny VCSELs will beam the data to equally tiny optical detectors 2 or 3 inches away. VCSELs will be able to transmit far more data than metal wires and they'll require much less power. This is just

the technology's applications. VCSELs will turn up in everything from printers to network switches. The first rollouts have already begun. VCSELs were

tscribed by Professor Kenichi Iga at the Tokyo Institute of Technology during the late 1970s. But it wasn't until 1989 that Jack Jewell at Bell Labs figured out how to make them practical. Jewell was a long-haired West Coast type in the staid confines of New Jersey and his employers saw his VCSEL project as too harebrained to waste equipment time on. So, when a friend at nearby Bellcore offered the use of his company's equipment one weekend, Jewell was finally able to fabricate the tiny lasers. Amazingly enough, they worked. The lasers J

ld crafted were incredible not just because they were small, but because they could be manufactured the same way as chips. That meant that many of the same techniques that have powered Moore's Law could now be applied to lasers. And they have been. Since 1991, researchers backed by Darpa and a handful of companies such as Honeywell and Motorola have refined VCSEL technology so that it is cheaper, faster, and more versatile. The results are jaw-dropping. Even if you

ty that hard, says Anis Husain of Darpa, a VCSEL is capable of transmitting data at 6 Gbps – and far faster rates have been shown in the lab. Unlike conventional lasers, VCSELs can be packed closely together to form two-dimensional arrays. That means a 10-by-10 array of VCSELs could be used for an aggregate transfer rate of 600 Gbps. VCSELs are a

by 10 by 2 microns. That's almost a hundred times smaller than a CD laser, notes Philippe Marchand, a researcher at the University of California at San Diego. And compared with a metal wire, which requires 20 to 40 milliwatts to transfer an equivalent amount of data, they require far less power to operate – a few milliwatts. Within a year, says Marchand, VCSELs will require only 1 or 2 milliwatts per gigabit per second. Suddenly, copper wires in a computer will be as obsolete as copper lines in the Internet backbone. We've alread

ehis transition with the copper wires that connect nearby computers. The newest generation of local-area networks – Gigabit Ethernet – relies on VCSELs and fiber-optic cables. Next, VCSELs will replace the ribbons that connect boards inside computers. The first rollouts, in the form of network interface cards, have already begun from companies like Vixel and Hewlett-Packard. Before VCSELs can really take off, manufacturing and assembly processes that allow for precision alignment between VCSELs and detectors on neighboring components will need to be developed. For now, fiber-optic cables are used as interconnects, but development engineers are working on ways to eliminate them. All of this

sd like science fiction. But just a few years ago, so did the thought of a Cray computer fitting inside a shoebox. VCSELs are now on the same dizzyingly exponential curve. By Jenni

S* | <st*

ITECH>  

parchOperation</a> <a href="htta

ip> <a href="htt

sneezso it's not always obvious when your machine has a virus. The remedy: a networked "immune system" in development at IBM. The software identifies a virus – old or new – on any PC and notifies IBM, where the research lab's machines analyze the virus, develop a vaccine that will prevent it from replicating, update the immune system's database, and send the vaccine back to the infected network. Within minutes, all the machines on the network have the code to fight that infection. Today, computer

stake months to spread, "but soon they'll be able to cross networks within days or hours," warns Jeffrey Kephart of IBM's Thomas J. Watson Research Center. "We need something that can respond faster." The immune-system software may be that something. The bad news is that IBM is in no rush to release this digital penicillin, saying only that it may be available sometime this year. By James Gla

e* | <strong*

T</stp> | <at

aive.ation</a> <a href="http://i

n/p> <a href="http://

i>Network com

or thilients, are stripped-down machines verging on dumb terminals, lacking hard drives or any real processing guts. Bean-counting MIS techies are presently invited to buy them en masse from Sun Microsystems or Network Computer Inc., an Oracle subsidiary – and replace all the hundreds of desktop PCs out there in cubicle land with something cheaper to manage. Translation: Goodbye

bert screensaverrewell uake deathmatch, sng duly legal JPEG collection. Thin clients are des

drun standardized, stripped-down apps – a browser, a wordprocessor, an email program – over a network, on a powerful server somewhere down in the basement. This will sound familiar to older kids who remember how the mainframe era kept computing power centralized in the hands of the gatekeepers. The MIS guys hated the PC when it first starting showing up in offices. After all, PCs are pain in the ass to install, maintain, upgrade, defrag, recover, et cetera. PCs are horribly individual creatures. A new upgrade across the company could take a year. But NCs don't need to be upgraded. Only the NC mothership does, and that could take mere hours. Ka-ching! But one man's fiscal

as another's worst fear. All personal files are stored centrally, so NCs are tailor-made for corporate surveillance. Further, the NC office suite applications currently available are moronic, crippled versions of their PC brethren. And forget about a game over lunch, or a wacky desktop pattern. Forget about taking your laptop to the park. Forget about everything except the cold hard fact of the job – which is exactly the point. Thin clients are cur

largeted at the bottom of the computing barrel – data entry, point of sale, and other fields that already call for dumb terminals. Let's make sure they stay there: mainframes suck, and something tells me that the good people don't want to go back there. By David Voss</e

p*| CRU*

<rongp> | <a hre/

iwiren</a> <a href="http://archw

itweezers ar

ke thearship Enterprise's tractor beon a microc scale. Ten years ago, Arthur Ashkin, Joe Dziedzic, John Bjorkholm, and Steven Chu at Bell Laboratories (now Lucent Technologies) demonstrated how to pick up and move tiny latex spheres using nothing more than a microscope lens and a low-power laser. Today, Chu, a recent Nobel Prize winner, is using tightly focused beams of light to capture and manipulate strands of DNA, as shown here.