Newsmaker: Bechtolsheim's machine dreams

Over a long and distinguished career, Andy Bechtolsheim has earned a reputation as a top-notch engineer. Now that reputation will be put to the test.

The task: Invent Sun Microsystems' next "hot box" out of the comparatively ordinary components of the x86 server market.

That's no mean feat. The server market for machines built with x86 processors, such as Intel's Xeon or, in Sun's case, Advanced Micro Devices' Opteron, is enjoying rapid growth, but it's hard to make one x86 system stand out above the crowd. What's more, Sun is entering a market full of rivals that have years of design and sales experience already under their belt.

The 49-year-old Sun co-founder nonetheless is confident Opteron will help Sun gain that edge. "If the world had not changed with Opteron, then Intel would still be building 32-bit (x86) chips, and it would have been too late for Sun to enter this market," Bechtolsheim said.

In 1995, Bechtolsheim left Sun to start Granite Systems, which developed 1-gigabit-per-second Ethernet networking equipment. In 1996 Cisco Systems acquired the company for $220 million. Seven years later, Bechtolsheim founded Kealia to build special-purpose servers for handling video. Last year Sun bought the start-up with the idea of migrating the Kealia technology to mainstream servers.

Sun hasn't been afraid to raise expectations for the Galaxy line of Opteron servers that now are on sale. "He's the most prolific and exciting and talented workstation and single-board computer designer on the planet," Chief Executive Scott McNealy said of Bechtolsheim when he announced the Kealia acquisition. "With this guy...designing Opteron servers, there ain't going to be nobody who has the class and breadth of computers we have."

Bechtolsheim spoke with CNET News.com about the difficulties Sun has encountered during its embrace of x86 servers, as well as the future of the Galaxy line.

Q: How is Sun different now than it was when you left in 1995?
Bechtolsheim: Well, the funny thing is that one of the suggestions I had when I was leaving was maybe Sun should consider building an Intel-type product line just to make sure that we had that part of the market covered. But Sparc was doing really well back then, and nobody had any interest in that. Now I've ended up doing what I proposed the company should be doing 10 years ago.

The point is that the company got a little too religious the last many years, prior to me coming back. It's a lot less religious these days. We obviously made a deal with Microsoft where we want to work with them to make life easier for customers to bridge the Solaris and the Microsoft operating environments. We're actually working with Microsoft on their services for management. And we now support the full range of operating systems, every version of Linux.

What did you do at Sun in your first career there? What led you out of the company, and then what led you back?
Bechtolsheim: Well, personally, I'm always driven by opportunities. We started Sun around the workstation opportunity (from) the work I did at Stanford--that's where the name came from, the Stanford University Network. Then Sun evolved into a server company, which was another great opportunity with the whole Sparc (processor) direction. In 1995, I saw an opportunity around changing the networking speed from 100 megabits to a gigabit. That got me very excited, so I left Sun to pursue that. I ended up being acquired by Cisco for a lot of money. I stayed there for the next seven years. The Cisco Catalyst 4000 and 4500 series was the product line that my group delivered to the market. It became the world's highest-volume modular chassis switch--I think they shipped over 50 million Ethernet ports.

We have nothing against Intel.

I got a little restless there a few years back, and I looked at some opportunities around media servers. This was when the Opteron architecture got announced by AMD and it was obvious to me that this architecture would make a significant difference in the market. Now you can't really start a start-up these days to be a server company. It's a little too late. The last two server companies to enter the market were Sun and Dell. It's really hard to enter this business on a grand scale. And so, as a start-up, I was looking at a video market as a vertical market segment opportunity.

When Sun announced that it was going to do Opteron servers, we connected and said we had all this Opteron stuff under development, and would love to do more of that at Sun. We very quickly came up with the deal that brought me back to Sun. Combining the team I brought with me with the existing people at Sun was really the first time Sun made an internal design commitment to industry standard (x86) architecture. Since my return in April last year, we've been very busy working--not just on the systems we're announcing next week, but on a whole bunch of systems that are not yet ready for announcement. All are based around the Opteron architecture.

Our original goal was to deliver a complete family of Opteron capability to the market. The first two members we're announcing are the 1U and 2U boxes (rack-mountable systems 1.75 inches and 3.5 inches thick), which is obviously the highest volume part of the market. In many ways the other systems are more interesting, but we can't talk about them today.

Yeah, I am very curious about, in particular, the eight-processor server.
Bechtolsheim: The company's public that it's working on systems up to eight-way. Obviously we are working on blade servers. But I just can't give you any more specific details on these systems.

So what led you back to Sun is that you wanted to do something like a server start-up, but not just for the vertical markets?
Bechtolsheim: AMD with Opteron is showing a lot of leadership doing the right things for the market and for customers. Building a new business at Sun around that was quite an interesting opportunity and really appealing to me personally. If the world had not changed with Opteron, then Intel would still be building 32-bit (x86) chips. It would have been too late for Sun to enter this market. You can't add any value to a market that doesn't change very much. People used to think with the Wintel duopoly--Windows and Intel--everything was quite stable and nothing would ever change. But now you've got AMD Opteron delivering the best server chip and you've got OpenSolaris and all the Linux stuff, so there's more competition in the industry-standard server space now than there has been in years. That makes life more interesting because it also means there's more innovation.

How much of the Kealia work carried on directly to the Galaxy designs?
Bechtolsheim: The more specific servers that were more media-targeted are actually not being announced next week. I know that they

More Newsmakers