SunLabs: Games of skill Sun bets on best minds to lead it into the future By Cate T. Corcoran

September  1996

At a time when other companies (Apple, IBM, AT&T) have or are looking for creative ways to cut their R&D budgets, Sun is increasing theirs. But that's probably because Sun Microsystems Laboratories is already so small it can only get bigger. With only about 100 people, it represents less than 1 percent of the company. Here's a glimpse at the expectations of SunLabs and some of the projects brewing there. (2,000 words)

Mail this article to a friend

Small and only five years old, SunLabs is not very prominent among research institutions. But the group has received a stir of recognition in the past few months. Ivan Sutherland, a SunLabs fellow and co-founder of Evans & Sutherland, recently won the Smithsonian-ComputerWorld award for lifetime achievement. "Dave Evans and Ivan Sutherland were pioneers in the creation of computer modeling and visualization techniques," reads the plaque posted in the lab headquarters lobby of Building 29 in Mountain View. And, of course, SunLabs claims credit for Java, although former Sun employees such as HotJava author Patrick Naughton, are quick to object to that claim.

Located on a narrow strip of land on the wrong side of the highway between one of Palo Alto's artsy-craftsy wooden welcome signs and an ugly gray slice of the bay, Building 29 is the designated birthplace of the future. Sun management allocates about $30 million a year to the lab in the hopes that it will produce another great moment in computer history, analogous to the invention of the graphical user interface at legendary Xerox PARC. But Xerox is also famous for not having made a penny from its genius, and like most research institutions these days, SunLabs has to justify its existence by being relevant to product development groups. That means turning projects into products that make money for the company. Current research areas include ATM networks, asynchronous systems, an operating system for clustered computers, and e-commerce.

Bob Sproull, a SunLabs fellow and vice president who oversees a lab outpost in Chelmsford, MA, is working on asynchronous systems. The idea is to design circuits that can communicate without having to run at the same clock speed. It's possible such systems could run faster, use less power, and be easier to build than conventional computers.

Another lucky five, including Berkeley Unix creator Bill Joy, work in Aspen on Spring, SunSoft's futuristic operating system.

Back in Mountain View, Distinguished Engineer Duane Northcutt is working on ways to better meld time-sensitive data -- video and audio, that is -- and distributed desktops.

"Punch-through video gave us effects, but you couldn't really do anything with it," he says, sitting at his desk and fiddling with 40 hours of news broadcasts stored on a remote server. "Velcro-ing a TV on the side of the computer is a much more cost-effective solution." He'd like to make it easier for people to use video on the desktop.

"For example, I'm writing a program that will recognize the color purple, so the computer can say, 'It's Barney!' and change the channel." Tilting his head sideways and looking up mischievously a la Princess Diana, he adds, "When you're doing this kind of work, you watch a lot of daytime TV, so these are real issues for us."

He wants to demonstrate the conceptual search concepts the Chelmsford group has been working on, so he brings up footage of the water balloon fight SunLabs waged against SunSoft a few years ago as a part of Sun's elaborate April Fools' Day prank.

"We only did it once because no work got done for about a month," he said. SunLabs built a mannequin piss statue, a 20-foot catapult to shoot water balloons, and a plexiglass minicar to keep ex-director Wayne Rosen dry. It also hired a helicopter and a water truck and broadcasted from the roof. "We're sleeping with your spouses, you'll never win..."

The video shows a troop of women in combat fatigues emerging from the building with water gun packs on their backs and black grease smeared under their eyes. The group includes Liz Kniss, Sun's outreach person and former mayor of Palo Alto, and Michele Huff, SunLabs legal counsel. Northcutt types in "Zander" (SunSoft's former leader) and finds all the frames with him in it. "The all-women's brigade gave Zander [quite a bath]," Northcutt explains. (Other SunLabs research revelations? "Normally attractive people look pretty heinous on a still frame.")

SunLabs won. "They outnumbered us, but we out-technologized them," Northcutt says.

"Tactics and style over brute force!" shouts a random pony tailed engineer, his arms raised in a victory cheer.

The SunLabs approach impresses my liberal arts major mentality. It drives home the point that engineering can have a real effect on people's lives, maybe more than ideas.

Products near, but not here
By its nature, a research group is somewhat removed from product development. But SunLabs has come up with tactics to put its ideas to good use. The lab's director, Burt Sutherland, prefers that we not think of SunLabs as an elitist institution where the brains work on great ideas and leave the dirty work of maintaining code and zapping bugs to the toilers in the product groups. Serious and possessing a military man's neat bearing (the Navy awarded him the Legion of Merit), he also happens to be Ivan's brother and was present at the birth of the GUI while he was manager of Xerox's Systems Science Laboratory. He points out that SunLabs sends about 15 engineers a year into product groups. In fiscal 1994, the lab sent two whole groups (Spring and SPARC) into Sun.

Other practices keep SunLabs tethered to the tarmac. It doesn't carry out basic research, nor do its projects involve technologies more than three or four years away from market.

Other groups in Sun are as likely (perhaps likelier) to shape our future. Think of JavaSoft's Java, SMCC's Internet servers, SunSoft's NFS -- or even the video about the office of the future that SunSoft produced last year. This playful corporate melodrama depicts an office where computers do the dirty work, everyone is free to think, each desktop boasts a giant screen, and computers are rented by the minute in airport booths. The curious thing about the video is that all the technology in it already exists. What's missing is the right economics for market acceptance.

The same could be said of much of what SunLabs works on. Much of its research is concerned with how to make good use of existing but high-end technology.

What distinguishes SunLabs from the rest of the company? The research group has a different risk profile and a different kind of accountability. At SunLabs, it's OK to make mistakes as long as the company learns something from it, Burt Sutherland says. In a product group, the last thing you want is a nasty technical surprise. It makes you late and over budget, he adds.

"The lab is really an intellectual trading post," to quote Sutherland again. It attracts talent who may not want to work in product group. Lab luminaries include John Ousterhout, a renowned Berkeley professor, distinguished engineer, and manager of the TCL/Tk group. New this year is Tom McWilliams -- a leading light in CAD for 20 years and founder of Valid -- who Sun hopes will help improve their internal chip-design processes.

SunLabs acts as a liaison with universities and other research groups, finding out the latest doings and reporting back to the product groups. This year the labs had 34 college interns in Mountain View. (The Chelmsford group was hosting a tour for interns from other labs in the area when I called. "I'm not wearing anything odd today," Sproull said. "I wanted to look respectable for the interns.")

SunLabs also coordinated the Java innovation program, an academic contest to create Java applications. The lab hosts at least three or four groups of visitors a day and conducts seminars about its research. Although the group doesn't give research grants, it arranges 40 to 50 university research contracts a year. SunLabs employees also visit customers so they can understand their needs and sometimes return the favor by giving a talk on lab projects.

"What have we done for you lately?"
How does management evaluate SunLabs' performance? It's a tricky question that all labs are wrestling with, as shorter product cycles put more pressure on R&D to deliver the goods. At Hewlett-Packard, 50 percent of all projects are expected to result in some kind of practical application, said Scrinivas Sukumar, director of Strategic Planning for HP Laboratories. McNealy doesn't hold the labs to any strict numbers, but they're supposed to help the rest of the company change its behavior, Sutherland said. Each year, the labs creates a report called "What have we done for you lately?" which summarizes the lab's achievements. Naturally, it's an internal-only document.

"There's a lot about Java along with patents and papers," Northcutt says while rifling through the latest version in his office.

Counting patents is one traditional way to measure a lab's performance, although it won't tell you whether or not a patent became a successful, revenue-generating product. SunLabs produces between one sixth and one third of Sun's patents, even though they have only 100 engineers compared with Sun's 2,500. In other words, five percent of the company produced 15 to 30 percent of the patents.

Compared with Mitsubishi, NEC, HP, and others, Sun has relatively few patents, but it also has far fewer employees and is much younger. HP Labs has 1,400 of the company's total 100,000 employees. Apple does not disclose R&D figures, but analysts estimate it spends about six percent of total sales (which were $2.2 billion in the quarter ended in June) on research, according to the Dow Jones News Service. That's approximately three times Sun's R&D budget. Sun's approach is conservative. The company spends less than one percent of revenue on research. The usual R&D budget -- for those companies lucky enough to still have one -- is something like seven to 12 percent of revenue.

Has the investment been worth it? Lab management points to all sorts of accomplishments, but it's difficult to pinpoint exactly what influenced what. "Sun is not one of the more visible research labs," says Richard Doherty, director of the Envisioneering Group of Seaford, NY, a technology assessment and market research group that has close ties to many high-tech labs.

Retroactive credit where credit is due
Java has become something of a political football, with SunLabs taking credit for the project and at least one of its original team members claiming that Java had little to do with SunLabs.

"We were put ostensibly under the SunLabs moniker," says Patrick Naughton, senior vice president and chief technology officer at Starwave Corp. of Seattle. "[But] we never really spent any time in the SunLabs building, nor did we fraternize with the rest of them. So SunLabs really can't take credit for Java at all."

The Green team, as the Java project was originally named, was created in 1990 before SunLabs existed. Eventually the team reported to SunLabs' then -- president Wayne Rosen. In 1992, the team was taken out of the lab and made into a wholly owned subsidiary of Sun called FirstPerson Inc., which was disbanded in the spring of 1994. Finally, Java was handed off to SunSoft and then spun off into JavaSoft.

"It is possible that the initial funding for Green came out of the SunLabs pot, but the group was set up separately in location and spirit from SunLabs up front," says a former Sun employee familiar with the project.

The miracle scheduled for Tuesday
Successful lab projects include Northcutt's 1994 research, which was incorporated into Solaris 2.3 and 2.4 releases to improve processor scheduling and system performance by giving users better control over the system. Dave Unger's work on the Self object-oriented language resulted in just-in-time compilers for Java. Jim Hanko's work on synchronizing disk I/O brought top-speed data transfer rates to Sun's video server. The Chelmsford group's work on distributed computing tools eventually became a shipping JavaSoft Java Workshop using remote method invocation.

The lab also designed a memory management unit with a flexible software page-table structure that became part of the UltraSPARC chip. Audio and video research (Project DIME) became ShowMe. And Spring became Project DOE, now NEO. The group also turns over its internal processes to the rest of the company. For instance, Sun has adopted the lab's information clearance process to protect its intellectual property.

SunLabs doesn't only develop technology. The group also studies how to use it effectively. The focus of the current project is how to measure the cost of computing; another is how to quickly and cheaply train employees using distributed video. (SunLabs is conducting test classes at Cal. Poly and Chico State.)

Such accomplishments don't sound as sexy as Java or the laser printer, but they're important. And you never know where the next great thing will come from.

"You can't schedule creativity," as Sutherland is fond of saying. While government investments in basic research and private investments in R&D are decreasing, there is also more venture-capital money floating around the Valley than ever before. Innovative products could come out of, say, an Internet start-up. Or from a company like Faroudga Laboratories of Sunnyvale, which licenses its ideas to other companies, which turn them into products. Or SunLabs.

Alternatively, all this attention to the bottom line could backfire. Whenever times are tough, U.S. companies (much more so than their European and Asian counterparts) cut spending on research. But historically, most of the important computer innovations of the last 20 years have come out of labs and most of them haven't made money for their backers because it was too early to produce them at affordable prices.

Ultimately, investing in R&D is a bit like playing the horses. "What I say to management is that it's largely a matter of affordability and confidence," says Sutherland. "It's an investment in the future, a gamble of that money on a set of people in the lab."

Resources

• SunLabs
  http://www.sunlabs.com/
• Evans & Sutherland Smithsonian-ComputerWorld award /960710/feature3/
  
• SunLabs
  http://www.sunlabs.com/research/
• SunLabs' Timeline of Project Successes /960201/cover/timeline.html
  
• Java WorkShop home page /developer-products/internet.html
  
• Spring home page /tech/projects/spring/
  
• Tcl/Tk home page /960710/index.html
  
• "Java: The inside story"
  /sunworldonline/swol-07-1995/swol-07-java.html

About the author
Cate T. Corcoran (cate.corcoran@sunworld.com) is a free-lance writer who grew up in Palo Alto. At the age of 11, she helped test Xerox PARC's mouse. Reach Cate at cate.corcoran@sunworld.com.

If you have technical problems with this magazine, contact webmaster@sunworld.com

URL: http://www.sunworld.com/swol-09-1996/swol-09-sunlabs.html
Last modified: