eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
In recent months, Sun Microsystems Inc. has made a big push into low-end, low-cost computing. The Santa Clara, Calif., company rolled out blade servers as part of its N1 data center virtualization strategy, as well as two low-end x86 servers, and promised to continue providing more of the same. But high-end Unix systems are still an important part of Suns overall strategy, and Clark Masters, executive vice president and general manager of the companys Enterprise Systems Products group, spoke with eWEEK Senior Editor Jeffrey Burt about Suns plans for its top-of-the line servers.
Low-end servers and blade servers have gotten a lot of publicity. What is Sun doing with high-end servers?
I think the high end matters more today than ever, really. At the $500,000-and-up price point—these are [International Data Corp.] data, not Sun data—in the year 2000, it was 20-some-odd cents out of every server dollar was spent on the half-million- dollar-and-up market range. At the end of 2002, that was over 30 cents, so that the amount of IT dollars going toward the high end … is larger today than ever before.
Is this because the systems are more expensive or because theres a growing demand for them?
Its two things. The weakness in the market we see is more the midrange.
So were seeing [high-end server growth] with server consolidation and data center consolidation and the drive toward efficiency. Also, were seeing strength in government spending, high-performance technical computing [HPTC], all of those things.
Whats driving the demand for the really high end?
Two or three key factors that I see. One is server consolidation. Two years ago, when I talked with customers, it was all about staying out in front of the wave. … It was the dot-com boom times. It was all about deployment.
Now, today, its all about doing more with less—total cost of ownership. How do I drive costs out of the system?
Another thing is, most large organizations are structured in business units, and a lot of business units have their own IT infrastructure, and now I think the political walls are broken down, that cost control is much more important than the autonomy of a particular business unit. You see people, to save costs, much more willing to consolidate workloads and combine computing environments, and that helps drive the high-end server business and data-center-class machines.
Regarding N1, can you provide me with an idea of how Suns largest servers—the 12K and the 15K—fit in with that strategy?
With N1, the better we can do at driving up the utilization and efficiency, the more applications we can dynamically provision. Thats a huge opportunity for us. So with the software tools were developing with N1, to manage and provision it, plus the virtualization in the hardware with domain and the Solaris operating environment, with resource management and software partitions—or containers—we have very powerful technologies to leverage, to simply be the best in the world at that.
How important is HPTC to Suns high-end computing strategy?
Its very important to Sun up and down the product line. … Were developing visualization technology like Java 3-D, for example. Thats big in the research and technical computing area.
Were finding that technical [computing] has much more growth potential and is becoming much more integrated with most every organization, whether it be manufacturing to do design optimization before you actually do implementations to biotech companies.
What are some of the other areas in HPTC that Sun needs to address?
Were very good at large physical memory, so that gives us an advantage. High-bandwidth I/O we have.
We have a storage business and very good technology there. When we get our UltraSPARC 4 machines—and I think in the worldwide analyst conference I said we would be introducing those before the next analyst conference, so about year-end or early part of next calendar year—that will have multiple threads … so it will double the floating-point performance that we have in the same footprint.
Long term, were investing in additional cluster technologies; investing in InfiniBand for high-speed networking, for both I/O and machines to machines; and also new processor technologies and interconnect technologies aimed at HPTC.