Network World: Research roulette

A LOOK INSIDE THE VENDOR LABS WHERE MILLIONS OF DOLLARS IN HIGH-TECH BETS ARE PLACED.

Building 29 on the Sun Microsystems campus looks nondescript; just another generic, hightech bunker divvied up into no-frills cubicles. The only tip-off that strange and exciting things may be happening inside is that meeting rooms are named after famous explorers and scientists, among them Darwin, Galileo and Magellan.

In one second-floor office, researcher Russell Kao is exploring the world of CAD tools for custom chip design. Kao has a unique perspective on his task. Because of a bad back, he lies flat on the carpet, feet sticking out toward the center of the room, keyboard propped above his waist, eyes staring up at a monitor mounted on a homemade wooden frame above his head.

In a small meeting room down the hall, principal investigator Neil Wilhelm, clad in bicycle pants and plaid flannel shirt, sits at a table discussing with two colleagues a project to embed a lightweight version of Java into handheld devices. At random intervals a disc-shaped rubber toy stuck to the table shoots up toward the ceiling. Wilhelm plucks it out of the air, slaps it back down on the table - in a spot not already occupied by the large rubber snake - and the conversation continues.

Elsewhere, researchers are poking around in esoteric areas such as asynchronous, or "unclocked," circuits; persistent objects; SD graphics; Web-based smart card readers; electronic money; multicast network protocols; and ATM network control platforms.

Welcome to Sun Laboratories, or SunLabs, where 100 of the most inquisitive minds from academia and the computer industry conduct research that they hope will lead to product breakthroughs three to five years into the future.

Similar to IBM and Lucent, two companies with large research organizations, Sun takes the doit-yourself approach to technology innovation. The company spends $1 billion per year on R&D, primarily to fund the efforts of 5,000 engineers who push products out the door. But Sun also invests millions (the company won't say exactly how much) in SunLabs, a small enclave shielded from the pressures of 12- to 18-month product cycles.

At the other end of the spectrum are companies such as Cisco, which also spent $1 billion in fiscal 1998 on R&D and plans to spend $1.4 billion in 1999. Cisco has no "pure," blue-sky research organization. Rather, when Cisco invests research dollars, it has a specific product in mind. The company relies on acquisitions to take the place of pure research; in effect, startups are Cisco's lab.

No matter how companies approach research, they all face tremendous market pressure to produce products faster. At companies with advanced research labs, this has meant a weakening of the traditional separation between basic research and product development. As Nortel Networks CEO John Roth puts it: "We've moved to a business model where research and development are no longer separate functions removed from the business. Everything that's being done has to go out to the marketplace, otherwise we shouldn't do it."

Playing it cool at SunLabs

At Sun, Bert Sutherland, who helped create SunLabs in 1990, says he has "fiercely resisted" having people on the product side dictate what projects the lab pursues. The advanced research group, which consists primarily of engineers but is also sprinkled with a sociologist, a medical doctor and a CPA, operates in a semiacademic atmosphere. It's a place where taking risks is the name of the game and where failure is an option.

In fact, Sutherland says he expects half of the projects to fail, although he's quick to point out there is value even in projects that do not result in marketable products. There is often some component of the research that can be used elsewhere, and the projects help the company avoid going down a technological dead end. "We make the company's technical mistakes early, while they're still cheap," he says.

For example, SunLabs made a foray into ATM but couldn't get it to work properly. So Sun's internal networks now use Gigabit Ethernet for high-bandwidth needs.

SunLabs' relationship to the rest of the company mirrors the atmosphere inside Building 29 - informal and loose. The genesis of a research project could be something as amorphous as CEO Scott McNealy telling Sutherland, "You know, security is a real problem that the lab should look into."

Projects are also launched in response to a request from a product group, a case in point being Kao's work on improved CAD tools. Ideas likewise come from the researchers themselves, as did the project to embed Java in handheld devices.

Or a project could be an investigation into one of Sutherland's pet areas, such as the economics of internal corporate computer networks. For that project, Sutherland enticed the controller of Sun's internal network, a CPA, to join SunLabs and lead the group. Sutherland's marching orders were: "Look into this area and be creative."

The group began looking into ways to impress upon employees the cost of various business resources, whether it's sending an e-mail with a large attachment across the world or attending an internal training program. The group came up with digital tokens, a form of smart card that is now being used in a pilot program inside SunLabs. For example, at the beginning of each year, each employee's digital token is programmed with a dollar figure equal to the training budget allotted to that employee. When the employee takes a training course, the course's value is deducted from the employee's electronic account.

The same research group was looking into authentication issues and came up with a Java smart card that can be used for secure access to a company's Web site. "That was an entirely unanticipated outcome of the work we started three to four years ago in economics," Sutherland says.

That's the beauty of a place such as SunLabs, where researchers have the time and freedom to pursue ideas even though they may have no idea what products will bubble up to the surface.

However, Sutherland emphasizes that SunLabs does not do the kind of pure research conducted on college campuses. "This is not an ivory tower where you come and retire as a fuzzy-haired scientist for the rest of your career," he says.

For a project to go forward, it has to meet certain criteria. "The first thing I need is a good, novel idea that's relevant for the company," Sutherland says. The researcher has to be able to answer the question, "How does it matter to Sun?"

Beyond that, Sutherland says he needs "a champion" to run the project and to recruit other people. "It's somebody who can stand up and say, `Yup, the next two years of my life are devoted to this one.' "

Finally, he needs a critical mass of researchers who have enough intellectual firepower to get the job done in a reasonable amount of time. Sutherland likes to have project teams of at least three people but not more than eight, and he prefers short projects over long ones.

Ultimately, Sutherland decides what projects go forward. The product divisions within Sun can make requests, but "the lab has very carefully reserved the right to choose the things it will work on." The reason is simple: If Sutherland has an expert who can take on a specific project, or if he is confident he can recruit one, he'll accommodate a request from a product group; otherwise, he just says no.

Take the ball and run with It

Once a team is up and running, Sutherland gives the researchers plenty of slack. Teams are required to submit progress reports, and Sutherland is always around for consultation, but he tries to manage with a light hand. "Sometimes people are doing things I wouldn't do, but lots of times I hold my tongue and don't criticize because a boss can pour a lot of cold water on a little tender spark of an idea."

Just as the method for generating and overseeing projects is relatively unstructured, so is the way research projects become commercial products. Rather than establish a formal technology transfer process, SunLabs relies on personal contact and events such as an annual open house to pitch its research efforts to the product groups.

And researchers have the opportunity to move into the product divisions along with their innovations. For example, Kao says he's prepared to move with the tool he's creating. "I'm looking forward to it. I want to see the stuff get used," he says.

About 100 people from SunLabs have transferred into Sun product groups over the past eight years, and many have also come back.

But SunLabs' arms-length relationship to the company appears to be changing. Shortly after Network World visited SunLabs, Greg Papadopoulos, Sun's new chief technology officer, replaced Sutherland as director of SunLabs and began a reorganization aimed at a tighter alignment between SunLabs and the product groups. Sutherland has become director emeritus at SunLabs and is an advisor to Sun's senior staff.

Papadopoulos combined the 100 SunLabs researchers and 140 advanced development researchers from the various business units into one team flying the SunLabs banner. He reorganized SunLabs into three areas that reflect the company's core competencies: networking and security, large-scale systems, and applications and services.

He also hired a director of advanced projects to add a marketing perspective to the research group. "We don't want to mess up the formula," Papadopoulos says. "We're just trying to bring in some market understanding."

A new age for Bell Labs

There are few places in which the struggle between advanced research and market forces plays out on a greater scale than at Lucent, which inherited Bell Labs in the 1996 AT&T breakup.

Founded in 1925, Bell Labs holds a total of 26,000 patents. Eleven Bell Labs scientists have won the Nobel Prize in physics, including Arno Penzias and Robert Wilson, who discovered the background noise from the Big Bang. Bell Labs is credited with pioneering work in transistors, lasers, digital switching, solar cells and cellular mobile radio. It developed the first 32-bit chip and the Unix operating system.

Under AT&T's paternalistic management, Bell Labs had the luxury of operating as a relatively independent research lab. Being thrown into a tough, competitive business has forced Lucent to make "big changes" in the way Bell Labs operates, says Melvin Cohen, a 34-year Bell Labs veteran who is currently vice president of research effectiveness.

For starters, Lucent management decided to limit spending on pure research to 1% of revenue, or about $300 million. That's less than AT&T spent, but Lucent only took 75% of Bell Labs, while AT&T retained the other 25%.

Under Lucent, the types of projects researchers work on are more closely aligned with the company's overall strategic direction. For example, Lucent quickly realized it lacked data networking products, so "we made a conscious decision to fill that gap," Cohen says. The result has been products such as the PacketStar IP switch, PathStar Access Server, Lucent Managed Firewall and LifeStar Optical Network System.

For the first time, researchers are required to leave their cloistered labs and visit with customers. And there is a tighter alignment between researchers and product groups, so "the transfer of technology to the business units happens faster," Cohen says.

When Lucent analyzed what the Bell Labs researchers were doing, they found that half of them were working on small joint projects with engineers from the product groups. Lucent reduced the number of joint projects to focus on just a handful, called breakthrough projects. Breakthrough products are those that offer the promise of "creating a big-time difference in the fortunes of that business unit," Cohen says.

The sky's the limit

That's not to say Bell Labs scientists don't do any blue-sky research. Work in biological physics, in which Bell Labs boasts the top department in the world, is ongoing. Cohen says the work may not result in products in the near future, but it's compatible with the lab's mission of advancing the understanding of how the brain works and how organisms communicate.

Real-world advances in those areas may not be that far off. Scientists at Bell Labs are developing chips that mimic the way the human brain receives and stores information, This research on electronic neural networks could lead to machines capable of visual perception and speech recognition.

Bell Labs is also working on systems that can translate information from one language to another in real time, telephones that can understand human speech, wireless data, superconductivity and optical fiber that can transmit a trillion bits per second.

Cohen says Bell Labs scientists have become "energized" by new entrepreneurial opportunities at Lucent, which simply didn't exist under AT&T. If researchers come up with new technology that doesn't mesh with Lucent's product line, there's a formal procedure for channeling that research into a new business venture.

Thus far, Lucent has created nine of these spinoff companies, including Inferno, which sells a network operating system that can link any device over any type of network; Elemedia, which sells Internet telephony software; Veridicom, which markets fingerprint authentication devices; Global Cast, which is developing multicast technology for the Internet; and Visual Insights, which does a type of data mining.

Common ground One thing research organizations of all stripes share is an academically rooted insistence on making their work public. Sun and Microsoft may be the most bitter of competitors, battling it out in court over Java, but both put their advanced research projects on the World Wide Web for the world to see.

Trying to keep research projects secret would be "a really short-sighted way of looking at things," says James Kajiya, assistant director of Microsoft REsearch. "When that happens, progress in the field peters out very quickly"

Sun's Sutherland adds that by publicizing its to tell us we're crazy and to find bugs in the software."

"By and large we run an open activity," he adds.As for the concern that other companies will steal SunLabs' ideas, Sutherland says, "it's up to Sun to eun fast. We have to count on the company's initiative and speed and nimbleness and flexibility" to turn SunLabs' ideas into the Javas of the future.

Weinberg is Network World's features reporter. He can be reached at neal_weinberg@2nww.com.