Interview: Sun execs talk server futures

Sun Microsystems made its name selling big, powerful boxes. But as the economy forced corporations to clamp down on spending, Sun watched its revenue slide with computer systems revenue down nearly 16 percent in fiscal 2003 compared with the year before. Buyers turned toward lower-priced, Intel-based servers, and in response Sun, which once shunned the boxes, rolled out low-end x86 servers of its own. Today, Sun is focused on offering more than hardware. At next week's SunNetwork 2003 Conference, expect to hear about packaged offerings from Sun that combine hardware, software and services to meet specific business needs. Jennifer Mears recently sat down with Neil Knox, executive vice president volume systems, and Clark Masters, executive vice president and general manager, enterprise systems, to hear about how the future of Sun's server business fits into its overall strategy.

Q: Sun has built its business selling expensive high-end boxes based on the SPARC microprocessor and running Solaris, your version of Unix. Yet in May you rolled out low-priced Intel boxes, and Scott McNealy talked about a focus on x86 running Linux. How do you see the server market evolving?

Knox: I know we've got a reputation of having high-end proprietary systems, but we've always had a very successful low-end business. In fact, the workstation business is still very large for us. What we've added to the SPARC low end, the US$2,000 category of server, is a set of x86 one-processor and two-processors systems running both Linux and Solaris on x86.

Q: There might be confusion about where Sun is heading with low-end Intel boxes and how that fits in with what you're offering on the high end. What can users expect from Sun?

Knox: The overall focus is solving customers' problems with technology, and customers will have specific ways that they're focused on. There's a vertical model where they really like the big iron, centralized mainframe in its capability, capacity, reliability, availability and serviceability. Then you have another set of CIOs who want to go on what they call a horizontal, scalable way, whereby they believe in smaller systems and having a multitude of smaller systems and giving themselves that reliability, availability and serviceability through that model. So to be a major player in the enterprise computer market you've got to be able to offer both.

Q: On the low end, how would you characterize your commitment to x86 vs. SPARC?

Knox: The commitment is strong in both camps. I mean we presented a business plan to the company in October of last year, which said we need to get an aggressive 32-bit, IA-32, x86 line going because that's what our customers are looking for along with a standard distribution of Linux. And we now support, or distribute, Red Hat and SuSE. And we're able to talk to the customer and say, 'Look, here's what you've got with Linux and x86: high performance, relatively cheap hardware. Not too secure, not too scalable, but it's quick development.' Compare that to SPARC Solaris Java with the industry's most reliable, most robust, most network-centric operating system, highly scalable. You have two very complementary product lines.

Q: How do you position the Intel boxes against the low end of the SPARC boxes?

Knox: Well, you show the customer the strengths and benefits and the features of both product lines, and we don't worry about the chipset. What we're really focused on is what's the application they want to deploy. What's the solution set that's really going to resolve their problem? We're non-religious about the technology that we offer.

Masters: Sun is one of the most open companies and supports customer choice. Every significant open standard in the industry we've invented: TCP/IP, (Network File System), Java, all these things. And if you look at other companies, I mean you could build a case that Red Hat is as proprietary as it comes: It only runs on Intel, and it's only compatible with itself. So please don't position us as proprietary. . . . That's exactly why we're doing Solaris x86. It's because now even our Unix operating system is on multiple platforms and through multiple suppliers, so again another statement for openness and customer choice.

Q: lanning to expand your Intel line, or are you focusing on one- and two-way?

Knox: Well, it's clearly the market sweet spot. Dell emphasized that (in July), they view the two-way as the sweet spot. It's where the volume is. We're keeping our options open at this point. If I have to add, I will. It's very much a function of what the customer base tells us. I have a highly scalable SPARC that goes in my lineup to eight-way, and then (Masters) takes over from there. So we believe we have the market covered, but we're keeping a close watch on it.

Q: What about server virtualization? What can users expect from Sun there?

Masters: We virtualized the hardware so we can take a big box and divide it into little boxes and then move resources around without interrupts. The next level of virtualization at the server level is containers; think of it as a software blade. The application thinks that it is running in its own dedicated server, but it's really just in a software partition, or blade, under one operating system, and resources are shared behind its back.

Q: Is that part of your utility computing strategy?

Masters: All those pieces are part of utility computing. Utility computing is one, a service-level arrangement between the customer and the vendor, a pay-as-you-go, pay-as-you-need sort of thing. We've got some of the first industry innovation with COD, capacity on demand, where we deploy hardware in advance and then when the customer needs it we turn it on. In utility computing, expect more announcements from Sun on bigger pushes on business strategies there.

Q: Going forward, where will you focus your research and development dollars? What do you see as the most pressing customer needs, and what new technologies will you introduce to address those?

Knox: At the low end, we're looking at solution sets of shipping complete integrated racks of multiple boxes as a single system. We're looking at, obviously, InfiniBand. We're looking at the blade environment as a very strong growth area, where we put a lot of R&D. Developing Secure Sockets Layer blades. Taking functionality perhaps out of the network infrastructure and putting it into the server infrastructure and again continuing that march to reduce complexity for the customer.

Masters: So if you look at the company level, where we're putting our money, roughly half the R&D, for software. Then second we have a big investment in SPARC. . . . We're investing in the platforms, both at the low end and the high end. We're investing in virtualization, some of the containers and the reabstraction of the relationships between the application and the actual equipment. We did the Terraspring acquisition, we did the Pirus acquisition, all within the last year on virtualization technologies, and we continue to drive it to our core. So all of this is about on every front, engineering out the complexity and putting in innovation. We're continuing to invest about $2 billion annually in R&D.

Join the newsletter!

Error: Please check your email address.

More about IntelRed HatSun MicrosystemsSuseTerraspring

Show Comments