Compaq aims for the top

Compaq (CPQ) fully intends to be one of the top three computer companies in the world by the year 2000.

John Rose, Compaq senior VP

By most estimates, the company is well on its way. Compaq recently shipped its millionth x86-based server, and has shipped more such servers than IBM, Hewlett-Packard, and Digital combined, according to International Data Corporation (IDC). Compaq is also the leader in shipments of corporate desktop computers, is close to gaining the top spot in consumer desktops, and is one of the top notebook PC vendors in the world.

However, it is the enterprise where Compaq hopes to continue to make inroads against Unix/RISC-based systems from vendors like Hewlett-Packard, IBM, Sun Microsystems, and Silicon Graphics. It is also the market that can turn Compaq from a $14 billion dollar company in 1996 into a $40 billion company by the year 2000.

John T. Rose is one of the most powerful executives at Compaq in charge of the company's most critical markets. As senior vice president and group general manager of the Enterprise Computing Group, Rose is largely responsible for spearheading Compaq's enterprise strategy, and consequently, Compaq's growth engine entering into the 21st century.

NEWS.COM talked with Rose about Compaq's views of the workstation market, Intel processors and alternative processor suppliers, the role of server-based clustering technology, and other topics in a recent interview.

Compaq recently made a major entrance into the workstation market with single and dual Pentium Pro processor-based systems running Windows NT. IDC forecasts the Windows NT workstation market will grow at 44 percent a year through the year 2000. Moreover, Windows NT workstations are expected to outship Unix workstations in 1997, IDC says.

NEWS.COM: Can you define what a personal workstation is? How is a workstation different from a $4,000 high-end desktop you can buy from a vendor like Gateway 2000?
Rose: You can't go by price range. If you understand the workstation business, you realize this is very much a vertical-market solution kind of selling. It is not [just] a hot box.

You can't just throw Fast SCSI and some memory and a network interface card in there. That does not give you a workstation. That gives you a hot-box product for some of those people in the gray area between a high-end desktop and low-end

PCs on steroids aren't workstations, Rose says

workstation, but that to me is not a workstation. You have to have a system designed for a balance as far as computing and network performance, manageability, cost of ownership, etc.

A lot of people throw out the statement that Gateway has workstations, or Joe Mumblefratt makes an announcement that they have workstations. Those aren't workstations; those are desktops on steroids. By the way, we sell lots of those [in the DeskPro line], but I don't count those as workstations.

So the workstation segment is like the server market?
Servers are following that path into the solution environment, and that's why we've focused on distributed enterprise solutions. Database solutions are different from applications solutions are different from back-office solutions. Maybe they're 90 percent consistent, but there is an element that's different.

You also have to realize that companies like Gateway are focused on the person who calls up and says, "I want one of those." We're focused at large opportunities in the hundreds and multiple thousands. These people would traditionally be buying RISC/Unix from Sun, SGI, HP, and DEC.

Are you finding that there are missed opportunities for sales because of a limited supply of Pentium Pro processors?
What we're seeing is demand that is large. I can say that Compaq has been driving Pentium Pro demand. We're probably the largest volume supplier of Pentium Pro and that's by design.

We're probably the most aggressive company out there. In the second quarter this year, according to IDC, we shipped 80 percent of Pentium Pro [servers].

The market changed over to Pentium Pro quickly, Rose says

Is it your understanding that the shortage is also due to the fact that the Pentium Pro is relatively complicated to manufacture and that yield hasn't been as high as it might be?
In any new product ramp, you're always challenged in one way or another, particularly in an environment where you create or change over a market much faster than you thought it could be. I think that's what we're all seeing.

How's your relationship with alternative CPU vendors like AMD these days?
Are there alternative vendors out there? [Laughing.] Alpha (from DEC) is in an economically unfeasible niche, and PowerPC is one of the historical screw-ups in the industry...Give recognition to HP for seeing the light and recognizing that the RISC architecture was not going to cut it. [HP has decided to work with Intel on the next-generation Merced processor.]

Are you still looking at alternative vendors like Cyrix and AMD?

They all tripped, and credit goes to Intel for driving frequency, volume, and cost. There are some areas where those companies will still have play, but they're not going to be a broad-based alternative to Intel.

I think there's clearly niche markets that are available for the Cyrixes and the AMDs. For example, in our Neteligent 8500 router, we use an AMD microprocessor. There are other areas; I don't mean to say they won't be there. They [x86 vendors] are going to be there probably more successfully than any RISC players that hang around. There's no room for them in our servers or workstations, though.

Can you describe what the transition to the next generation of processors like Intel's 64-bit Merced looks like? [Merced, sometimes referred to as the P7 processor, will be Intel's first 64-bit processor and is expected to offer a dramatic increase in performance over current 32-bit processors like the Pentium and Pentium Pro. Intel is expected to introduce the chip in 1998, or 1999 at the latest.]
Across all products we have aggressive technology road maps. With Merced-class workstations and servers, we will drive for a leadership position and drive the transition [to new technologies] hard.

Merced is very key to our long-term leadership in workstations and servers. With that class of 64-bit capability, we have the finishing touches into the high end of where the minicomputers and mainframes have been.

Clustering technology also takes you into the very high end of computing. What are some of the trends in clustering and symmetric multiprocessing (SMP) that you see? [Clustering allows servers to be connected to each other via a high-speed interconnect, providing a high level of fault tolerance, and, in the future, scalability in performance as more clusters are added. SMP, on the other hand, refers to the number of processors inside a single server, which can then be clustered with other servers.]
Clustering four processors offers a tremendous breadth of capability as far as various applications that can utilize that environment. Eight-processor [SMP systems are] certainly feasible, but that is probably not as large a market segment as clustered four-way systems. Certainly there are some applications you could envision coming off of the legacy environments and going on eight-way [SMP] systems.

It's reasonable to expect that we will have eight-way [SMP systems that can be clustered together]. Then again, what you'll see is a continuum on themes like overall performance, input/output performance, asset management, and cost of ownership.

Rose on the advantage of using Wolfpack API

You'll see clustering in 1997 and 1998 even prior to Merced challenging a lot of mainframe systems with Pentium Pro and next-generation processors. There is a logical evolution between Pentium Pro and Merced that advances 32-bit performance.

What's the importance of the Wolfpack API ?
We've invested a lot in working with Microsoft in evolving Wolfpack. There is a huge momentum that you're seeing in the industry towards NT, both in the desktops, in the workstation environment, and, of course, in the server environment.

If you look at the rate of change that is occurring around NT, it is logical to see that extension into clustering and high-availability systems as continuing to be along that trajectory of NT. It's surfing a wave.

Rose says Compaq needs to expand service offerings for enterprise

We recognize that as you expand more into the enterprise, there are enterprise customers that are looking for total and complete accountability and a comprehensive set of services. What we are doing is working with channel partners, particularly those who are adding expertise to do consulting services and presale/postsale services, and more system-integration types of activities.

As we move more into solutions, solutions consist of not only hardware software, databases, applications, etc., but also service is a key ingredient of that as well.

We're not necessarily going to deliver service ourselves but will have service products and a delivery mechanism that is tied to the high-response expectations that the customer has. We're not out to hire and deploy tens of thousands of people on the street like the legacy players have. For the most part, we haven't seen requests from customers to do that.