Don Harbert Heads Unix for DEC
Name: Donald Z. Harbert
Position: Vice President, Unix Business Segment, Digital Equipment Corp., Maynard, MA
Years in the Industry: 24
Years in Current Position: 2
Place of Birth: Seymour, Iowa
Family: Married 25 years to Peg, a school teacher; a son and a daughter, both in college
Car He Drives: 1985 Volvo 850 Turbo
On the Common Desktop Environment: "My personal view is that CDE is going to be overwhelmed by browsers. There are certainly going to be some Unix people that use CDE but very soon we're all going to be looking at browsers, whether it's from a Unix desktop, a Macintosh desktop or a PC desktop."
On the World Wide Web: "The Web has given Unix as an industry new life; there's no question about that. It's the first time in my memory that this kind of a wave was based on a software standard--TCP/IP--which was funded by the government but has always been a major attribute of Unix operating systems. Other operating systems are working very hard at adapting themselves to the Internet environment. But I think it does give the Unix vendors a window of opportunity to reestablish themselves in a new way as this market evolves."
Don Harbert is the top Unix executive in one of the world's major vendors of computer hardware. Although he quickly brushes aside any notion of himself as a guru--he's an electrical engineer and not a programmer by training--he's clearly in a position to influence the future of open systems. Digital Equipment Corp., though lacking the aura of a pioneering open systems innovator, now reports sales of more than $1 billion a year and more than 50 percent growth in its Unix business over the last complete fiscal year. On a straight revenue basis, DEC's proprietary VMS operating system still accounts for more of the company's business, but VMS and Unix are now about equal in its sales of first-time systems (rather than upgrades). Windows NT, the company's other supported operating system, accounts for much less business than the other two but is growing fast. "There's no question that we can't make the corporate business plan in this decade unless Unix grows at 30 to 50 percent a year," Harbert says. "That's the plan that the company put in place and it's a challenge. We've been delivering on it for the past two years and we have to deliver on it for a few more."
Harbert was born and raised in Iowa, in a town that then had 1,200 people and now has dwindled to about 600. His career choice seemed to come naturally: "For some reason, I always wanted to be an electrical engineer," he says. "I don't remember when I made that decision. It seemed like electronics and electrical engineering were going to have a very bright future and offer a career that was going to have a lot of challenge and opportunity."
"I decided I liked the smaller machines better than the big machines."
After earning his bachelor's degree from Iowa State University in Ames in 1970, Harbert went on to get his master's there in 1972, both degrees in electrical engineering. Along the way his career goal became a little less lofty: "I was thinking I was going to be a solid-state physicist but I had a little trouble with quantum mechanics," he remembers. Along the way, he had a chance to work using a DEC PDP 11 as a communications processor in his graduate project. "I decided I liked the smaller machines better than the big machines," he confides.
As a Vietnam-era student, Harbert chose Air Force ROTC as his military option. After college, he served five years on active duty, all of it in Southern California with the Air Force's Space and Missile Systems Organization, working on satellite and missile development with military contractors. After his Air Force duty, Harbert spent another four years in El Segundo, CA, working for a 25-employee military contractor called Questron, directing office operations. "I did everything from preparing contract proposals to fixing the Xerox machine," he says.
Move to Digital
By 1981, he and his wife had grown weary of life in La-La Land. "I found Southern California to be too crowded and too self-centered, or something," Harbert says. "Everyone was into their own thing. There was no feeling of any kind of community." He landed a job with Digital's semiconductor group in Hudson, MA, where he worked for 18 months managing VLSI design. Then he moved to the VAX systems group, working on advanced computer chip development and combining his talents in electrical design, software engineering and project management. In 1983, as manager for system technology in the Mid Range system group, he started the first design group outside the company's semiconductor group to do custom large-scale integration. "The circuit design level of VLSI is one of the few places left where there's serious electrical engineering still being done," he says.
Moving up in levels of responsibility, in 1989 he became head of the Mid Range system group, managing development of the VAX 6000 system project, the first VAX symmetric multiprocessing machine built by Digital. "I had a continuous career path that moved from the component level to the systems level," he observes.
"He oversaw the transition to Digital's Alpha chip, its pioneering 64-bit architecture."
Harbert got his first software job in 1990, as manager of the VMS operating system group. As such, he oversaw the transition to Digital's Alpha chip, its pioneering 64-bit architecture. "The major engineering work we were doing in VMS at the time was moving the operating system code from the VAX architecture to the Alpha architecture. It was my responsibility to figure out the planning, resources, schedules, and the risks we were going to take around porting that code. And to some extent, I had to keep the group motivated. For some people it was a tough time moving through that architectural transition."
He also had to deal with issues surrounding the long-term future of VMS, the company's proprietary operating system, such as how much functional content to make available on the first Alpha release. "We knew we had to have enough function so that the leading-edge customers would be willing to buy and start using Alpha," he says. "Also, we had to decide if and when to do cluster capability that adapted to both the VAX and Alpha architecture. How much resources would we dedicate to the Alpha project? At that time all the revenues were based on the VAX product. We had to make a decision on when to quit doing new VAX releases and go fully to Alpha."
For a period of months, Harbert headed Digital's system software group, when the company combined its VMS, Unix and Windows NT operating system organizations. Then, in a major corporate reorganization in April 1994, the groups were split and Harbert took charge of the Unix business segment, the job he still holds. As such, he's in charge of not only Digital Unix but a compiler group and Unix-related objects and messaging middleware.
Harbert summarizes Digital's current Unix activities as follows:
Operating System: "We have the continuing evolution of the Digital Unix kernel. We're making a major program investment in Digital Unix clusters for availability, scalability and manageability. We also will continue to make a major investment in storage management tools and products such as the advanced file system, and hierarchical storage management tools and technologies. We also have a major investment in scalability to support Alpha performance in a symmetric multiprocessing environment. Right now, we support 12 processors and 13 gigabytes of memory; we have announced the intent to go to 28 gigabytes of memory. So we have to worry continually about any kind of limit the operating system might impose on performance as we scale both processors and memory size and I/O as well."
"We're making a major program investment in Digital Unix clusters for availability, scalability and manageability."
Compilers: "The primary goal is to ensure that the compilers deliver all the performance that the Alpha chip is designed to execute. That means we do major investments in the back end of the compiler. We have three compilers that we invest in rather heavily: C, C++ and Fortran, all of which use the same back end. The back end is where we have to ensure that the multi-issue aspects of the Alpha chip get utilized to the maximum that the hardware allows. We have a parallel Fortran that we think is recognized as one of the two best in the industry. We ship C with the operating system. C++ is based on a lot of the same code."
Middleware: "We have a product called ObjectBroker that implements the Common Object Request Broker (CORBA) specification. It was among the first products to be released implementing CORBA. We have an asynchronous messaging product called DEC MessageQ--that certainly competes very favorably with IBM's Message Q product--that we sell to several large customers that are doing in-house development."
"Digital's 64-bit Alpha architecture, as well as its work with clustering, will have the largest impact on the future of the industry."
Harbert feels that Digital's 64-bit Alpha architecture, as well as its work with clustering, will have the largest impact on the future of the industry. The 64-bit programming model pioneered by both Digital and Silicon Graphics has been accepted as part of the Unix 97 specification, he notes, and has already made an impact. "We certainly have more experience with clusters, through our VMS experience, than virtually anyone else in the industry," Harbert says. "So we're trying to bring all that experience--in an engineering sense and in a field support/customer support sense--to Unix. We want to provide to our Unix customers the same kind of availability and scalability that VMS customers have enjoyed. Clusters in general will have a big impact on the Unix industry. We're seeing that now in the announcements from Sun and with continuing announcements from IBM and HP. Our technology will continue to provide some pioneering work there. On the scalability side, we hope to continue, in conjunction with our Alpha hardware partners, to set the benchmark for performance."
Back | Table of Contents | Next