Embedded Systems: Jerry Fiddler
Change is music to his ears

It's a good thing that 47-year-old Jerry Fiddler has Wind River Systems Inc.-the company he started in 1983 with David Wilner-to keep him busy. Otherwise, by his own admission, Fiddler would be jumping from job to job. It is not a matter of ability, determination or stick-to-itiveness. Rather it's a low tolerance for boredom and an itch to move on to the next challenge.

Fiddler is seldom bored and often challenged at Wind River Systems. The Alameda, Calif., company began as an embedded-systems design consulting firm, evolved into an operating-system supplier and now ranks as a full-spectrum supplier of tools, OSes and services. In the company's 15-year history, Fiddler has been, variously, the chief and only programmer, then-after a flip of a coin with Wilner-president and chief executive officer. Now he is chairman of the board.

As Fiddler's job titles and the nature of the company itself have shifted over time, so has the industry to which it belongs. Embedded software is no longer a monolithic domain populated by large companies with large problems to solve. Instead, numerous small- to medium-sized companies serve a variety of industries, such as medical systems, telecommunications, industrial control and consumer electronics. Now the industry is expanding into the exploding field of embedded consumer computing and Web-enabled devices, crafting software for such things as set-top boxes, smart phones and information appliances.

"I certainly haven't been bored," Fiddler says. "I've had more than enough to do, and I have enough challenges to keep me from being bored for a long time."

Professional challenges aside, if there has been any one constant, consuming interest in Fiddler's life, one from which he has never wavered, it is music. Indeed, the love of music has been behind many key decisions concerning the direction his life has taken.

In 1974, after graduation from the University of Illinois with a double major in music and photography, Fiddler's direction seemed pretty well set. But while playing electric guitar in a fusion band called Afterbirth, he saw a job offer as the resident composer for the University of Illinois' dance department-a gig he really wanted. To be eligible, Fiddler needed to go back to school and he did, choosing to get a master's in computer science because "it looked interesting" and coincided with an interest in computer-generated music.

In 1977 Fiddler gave up the idea of becoming a professional musician. He was living in Urbana, Ill., and playing guitar in another band, called Random. In search of a new direction for his life, he found it in a paper on voluntary simplicity out of Stanford Research Institute in Menlo Park, Calif.

Aha! Fiddler thought. "A think tank. I think I'll go there and look for a job." So he hopped in his '65 Dodge pop-top van and drove to California. There was nothing available at SRI, but a bulletin board there directed Fiddler to the University of California's Lawrence Berkeley Laboratory, which needed somebody with both a humanities background and computer know-how to build human interfaces to big systems.

Fiddler's work at Berkeley Lab was to write software that enabled computers to control things-physics experiments, say, or big linear accelerators, or a neutral beam source to fuel a fusion reactor. The job taught him not only about the technicalities of control systems, but also of the need for systems that operate in real-time.

The work also required him to write programs by employing coding conventions-essentially, style guides that make it easier for programmers to read each other's code-and modules, pieces of software that can be reused as needed. At the time, such techniques were novel; software was still generally written line by line, not chunk by chunk. Fiddler says the Berkeley Lab work gave him the ability to approach problems more broadly: Instead of trying to solve a physics problem itself, he learned the importance of stepping back to determine the pieces needed to build a system to solve it.

But that way of thinking about programming and solving problems only reinforced what he had learned in his undergraduate training in music. "Conceiving and scoring a musical composition involves many of the same processes that you need in programming," Fiddler says. "Looking at what you want to accomplish with a piece of music, breaking down the problem into smaller pieces to score, and linking them into a coherent whole to achieve an end." In a sense, he says, composing music is just another form of programming: only the language, the syntax and the rules of composition are different.

By 1981, Fiddler had grown bored and frustrated with his job at the lab and was itching to get back to the passion that drove him toward computer science in the first place: computer music. At the time, composing on a computer meant buying a minicomputer, building a synthesizer and an interface, and writing the software to make the computer play the synthesizer.

Fiddler duly acquired a PDP-11 computer, quit his job and, to pay the bills, hung out his shingle as a real-time systems consultant. His first customer was Control Video, a videotape-editing company. That is where he heard a sound as sweet as any he produced with a musical instrument.

"I had been working for days on making the application work and at about 2 a.m. I tried again, with no expectation that what I did was going to work," Fiddler recalls. Then he heard a "click-click-whir" of the recorder responding, "and I felt a sense of exhilaration that I can remember to this day. It was a life-changing experience."

This work led to a string of projects in the video industry, including a real-time videotape duplication system, a high-powered video editor for the National Football League and a special storyboard prototyping system for director Francis Ford Coppola.

Several months later, Fiddler asked his former Berkeley Lab colleague David Wilner to cover for him during an extended vacation. When he returned, the two men formed a partnership that resulted in the founding of Wind River Systems. The company was incorporated in 1983. Each partner contributed $3,000 worth of computer equipment and a desk, which together made up the total capitalization of the firm.

Initially, Wind River specialized in consulting on real-time software for large, complex applications. In 1987, the company responded to the growing demand for its services by becoming a software manufacturer, packaging its operating system and development tools into an industry-leading product called VxWorks. That real-time OS has matured over the years to fit into many environments, from very large applications, like telecom switches and aerospace control systems, to very small ones, in standalone, deeply embedded appliances.

Fast-forward 10 years, and the challenge that's facing Fiddler today is managing the growth of his company in an embedded market that's bursting at the seams in terms of revenue and possibilities. One of the most serious is the dearth of programmers with a detailed understanding of the underlying hardware architecture for which they are developing software. This lack is exacerbated by the emergence of powerful 32-bit embedded processors with capabilities that dwarf those of even five years ago.

"We are getting programmers who may understand a particular application such as digital cameras very well, or a particular software environment such as Windows," Fiddler says. But ask them about the nitty-gritty of embedded devices, such as race conditions, deadlocks and interrupt latencies, and "all you get is blank looks."

Fiddler is looking to address this problem in two ways. First, he is shifting Wind River's direction to developing tools that are more friendly to programmers who have come out of the desktop world. Second, he is evangelizing, trying to change the way computer-science graduates are educated. "We have raised a generation of programmers who think it is okay to press the control-alt-delete [reboot] keys once in a while," he says. "People are willing to make a lot of compromises when they are dealing with what is obviously a computer, but not when they are operating a home appliance or driving a car."

Net challenge
Another challenge will be to determine just how the Internet will change the nature of embedded systems, and what Wind River has to do to participate. "In a few years, most of the devices hooked up to the Internet will not be desktop computers, but all sorts of embedded devices," Fiddler says. "Once you put an embedded system on the Internet, you change its nature fundamentally. Now it is in a much more open, less predictable environment and it needs to be reprogrammable and field-updatable. This all changes how you design and structure a system and how you program it."

Beyond such issues, Fiddler is still very interested in music. Last year he recorded and produced a CD called "Xaz." If and when he runs out of challenges at Wind River Systems, Fiddler still has his music. He always has the option of quitting, putting together a band and going on the road.

Return to: TimesPeople 98