As vice president and general manager of Dell Computer Corp.'s enterprise server division, Randy Groves is among those shepherding the PC maker's push into the high-end server market with Intel Corp.'s Itanium architecture. In this interview with Computerworld, Groves talked about the state of the server market and Dell's plans for it.
Q: What do you see as the broader trends today in the enterprise server market?
A: As Intel servers get more and more optimized for the data center, we have seen a huge shift [in demand] towards rack-optimized and density-optimized servers for those environments.
There [also] seems to be two approaches to scaling going on in the industry. The more traditional approach is called scaling up, which means going to bigger and bigger symmetrical multiprocessor servers. You put more and more processors in one server and run one copy of the operating system.
The other approach is to scale out, which is done by adding computers [as capacity needs increase]. Web servers are a perfect example. You can very easily scale the load by adding servers and reprovisioning the work.
Q: How is Dell responding to these needs?
A: We are going after a "bricks-and-blades" approach for future servers. The blades approach is really a way to put a server on a card. It specifically reduces costs for customers by letting them put more servers in less space with less power and in an environment that is easier to manage and easier to reprovision. The focus is on reducing the total cost of ownership.
The bricks architecture gives you a flexibility that you've not had before in the back-end server space. The [architecture] allows you to build systems in a modular fashion using bricks for I/O, bricks for memory, bricks for storage. It allows you to mix and match bricks to get the right mix of processors, memory and I/O that you want.
Q: How are enterprise users' requirements different from those in the lower end of the server market?
A: The demand for uptime increases and the criticality of the applications [that run on Dell boxes] increases with every step we take up into the enterprise. From a hardware perspective, we are meeting the availability and reliability requirements. We already offer [99.9 percent] reliability. The vital requirement really is to make the software bulletproof, so that you can have the [99.999 percent] reliability that is needed in the enterprise.
Q: What specifically does Dell bring to the enterprise server market?
A: Dell's fundamental strength in this whole space is how easy we make it for customers to deal with us. When customers deal with us in a server environment, they realize how much they can save in being able to set up and configure their servers exactly how they want it.
Q: What does the arrival of Intel's Itanium mean for users?
A: The computational capability of Itanium is staggeringly better compared to IA-32 processors. It provides a very significant performance advantage for scientific applications and applications such as cryptography. But the most important attribute of Itanium is its 64-bit addressing capability. It allows servers to address more than 4 GB of memory [a capability needed to run large applications such as databases].
Q: Do RISC/Unix vendors such as Sun Microsystems Inc. and Hewlett-Packard Co. have an advantage because they, unlike Dell, make both the server hardware and software?
A: They have an advantage today. But there is nothing technically that will keep Windows and Itanium from getting there. There have been vast improvements in Windows, and there are some very impressing uptime numbers from Microsoft. By the time we get to Whistler on an IA-64 platform next year, the environment will have reached that necessary threshold.
Q: Are commodity Wintel platforms really suited to meet the specialized needs of the enterprise server market?
A: I remember having similar discussions around Unix and RISC a few years ago. We are seeing the same trend happening here again. Intel and Windows have taken over the workstation space and have begun doing the same in the server space. The fundamentals are there. It is going to happen. The reason is the economics behind it. When all the research and development are focused on one platform and the costs can be amortized across the industry, it reduces costs for users. It will happen when there is no benefit to a more expensive solution.