Vendors may offer to "help" potential buyers with the TCO calculation, perhaps taking the opportunity to twist the equation so it emphasizes their lines' unique features. But TCO is one of those intangibles that make evaluation difficult, if not impossible. Procurement, upgrades and maintenance costs are relatively easy to figure. Other costs, such as staff time, utilities, floor space, procedures and planning, are harder to factor. Much of what's done to a server through its life cycle is prescribed by the applications for which it is used. The application cycle may drive different usage patterns for servers and, in some cases, render what seemed to be impressive TCO-inspired options relatively useless. For example, hot-memory upgrades are great for nonstop applications, but in other instances they can be replaced by good planning. So does that feature lower your TCO? Typically, the answer is, "It depends."
Make no mistake--the vendors are in a tricky position. Margins on servers are razor thin and the economy of the past two years has led to very little growth in any market segment. Gartner says the overall U.S. server market was $55.5 billion in 2000, it shrunk to $48.5 billion in 2001 and won't see $55.5 billion again until 2005. What growth there is in servers often reflects buyers moving away from some of the high-margin, high-priced systems toward lower-cost, more modular systems. Gartner says it expects mainframe revenue to fall 8 percent by 2005, with very small increases in each category from high-end servers down to entry level. This indicates that customers are looking for more value in cheaper form factors. In one sense, the vendors have been their own worst enemy: They've finally found themselves on a path that offers price-performance improvements faster than corporate America's appetite demands.
IBM: pushing the Total Solution
IBM's Intel-based rack-optimized systems are known as its xSeries servers. In talking with IBM, we could tell instantly that its x-architecture was just a piece in a much bigger puzzle. Significant shares of IBM's servers are sold along with other IBM systems, software and services, and they are designed to operate best in that environment. The benefit to IBM customers is clear insomuch as management and fault-tolerance/avoidance methodologies are similar throughout the company's server products.
For those considering IBM servers but not other IBM software or systems, the benefit is less clear. While the technology that IBM adds to what would otherwise be a generic Intel-based server is sound, incorporating it often can put IBM on a different product-release schedule from Dell or HP. Customers may also find that taking advantage of IBM's best server tricks is expensive, as IBM says it will unbundle high-end management features soon. It should be noted that IBM's wares are price-competitive: HP's ProLiant boxes usually arrive with the largest price tag. However, separate management features will force enterprises to build a separate, costly management operation just for the servers.
In terms of time to market, IBM doesn't necessarily aim to be first. As an example, HP already offers a wide range of blade-based systems. IBM will be releasing its first system about the time you read this. To its credit, IBM took the opportunity to soup up the performance of its blade offering, using Xeon DP (dual processor) technology rather than the low-power Pentium III chips used by Dell and HP. IBM also plans high-performance Ethernet and Fibre Channel switches for its blade boxes. IBM sees some of the 1U market going to blades, with enterprises using blade systems for core applications--e-commerce, clustering and line of business applications, for example. This is in stark contrast to HP and Dell, which tend to view blade systems as edge devices, more appropriate for serving Web pages and lighter-duty tasks, such as DHCP and DNS services.
On the management side, IBM's eLiza is the company's overarching project aimed at putting system administrators out of work. ELiza promises self-healing, self-configuring, self-administering and self-updating servers that do everything system administrators do except eat donuts and spill coffee. On xSeries machines, mass configuration, dynamic partitioning, automatic updating, memory mirroring, real-time self-diagnostics and other features offer real advantages to users and are available now.
RSS
