Tech Industry

Commentary: Microsoft's Datacenter may signal new race

Triggered in part by Microsoft's development of Windows 2000 Datacenter, IBM, Hewlett-Packard and Compaq all announce plans to market 16- and 32-processor Intel-based servers.

Triggered in part by Microsoft's development of Windows 2000 Datacenter, released yesterday, IBM, Hewlett-Packard and Compaq all have announced plans to market 16- and 32-processor Intel-based servers.

This will allow Intel/Windows 2000 servers to compete in performance with high-end Unix servers sold by these same companies and by Sun Microsystems, which was notably absent from the list of Wintel server sellers. Instead, Sun today is announcing the availability of its first servers based on the UltraSparc III chip.

See news story:
New era dawns for Intel servers

Windows 2000 Datacenter is a truer substitute for Unix and other "big system" operating systems, but Datacenter will not turn the server world into a commodity business. High-end servers require customization and support, and they tend to accumulate custom interfaces over time as users adjust them to changing business needs.

Customers depend on the companies they buy from to provide on-site support to guarantee that the servers will continue to operate. That costs money, but businesses are willing to pay that cost.

Consequently, the arrival of Datacenter will change the nature, but not the intensity, of server competition. With one obvious exception--Sun--the same company that dominates Unix sales will dominate Datacenter sales. Ultimately, the costs of server-class support will be more than component costs, which will lead to pricing parity between Unix and Datacenter.

How will the nature of competition change? Behind Datacenter, server vendors over the next 18 to 24 months will scramble to introduce, at the low end of the market, packaged "server appliances" comprised of common hardware, operating system and middleware, with limited or no on-site support. These are appropriate for tasks that require no customization, where they can be expected to run without interruption for years--for instance, serving up Web pages and file-and-print operations.

What makes this interesting is that the server appliance represents the first huge shift in low-end server market dynamics since Novell popularized the file-and-print server.

Will Intel dominate the appliance part of the market? Absolutely. Will the market choose Windows or Linux? That is an interesting question. We think it will be Windows, but Linux is also a possibility.

The company most negatively affected by Datacenter is Sun. Because Datacenter is a strong substitute for Unix, Sun will no longer be able to demonstrate clear technical differentiation between Solaris/SPARC and Wintel--which may explain its recent move to purchase Cobalt Networks (an Intel-based server appliance supplier). Sun will be put under enormous pressure to conform to the server appliance standards that will be defined in the Datacenter world.

Although we don't expect any credible new entrants in the general-purpose server market, two companies currently "on the periphery" will be very tempted by the opportunities presented by Datacenter: Dell and EMC.

If Dell wants to play in the broader server market, it must develop a viable complex server support infrastructure. Until now, Dell has avoided investing in the support resources necessary to sell bigger configurations of hardware.

For example, the company has not made major inroads into the enterprise storage arena, arguably a simpler domain than servers. We do expect Dell to bring its direct sales model to the server appliance world. We don't expect them to invest in world-class complex server support.

EMC may be tempted by the server appliance arena, with servers essentially positioned as peripheral to storage. To date, EMC has demonstrated no interest in the server arena, but as storage dominates total server hardware spending, which is projected to surpass 75 percent by 2002, their competitive position will improve.

For IT buyers, the entry of Wintel into the high-end server market simply offers an alternative technology. Users should avoid becoming involved in the Unix-Wintel religious wars and concentrate instead on the needs of their applications and environment and on their sourcing strategies and long-term relationships with their sellers.

If Wintel offers an equal or better solution at a comparable or lower cost with the same service guarantees from established sellers, then users should consider it on that basis and not because the box has an "Intel Inside" label or runs Windows rather than Unix. META Group analysts Peter Burris, William Zachmann, Dale Kutnick, and Val Sribar contributed to this article.

Entire contents, Copyright © 2000 Meta Group, Inc. All rights reserved.