Hardware Today: Getting Blood From a Stone, Maximizing the Server Budget
December 2, 2003
Maybe it's a case of too much punch at the holiday party, but did you just hear your CFO mention the word "economy" without the requisite wailing and gnashing of teeth that accompanied it for the past two years?
Perhaps. Forrester Research saw a conservative 1.7 percent growth in IT budgets during 2003, but the research firm predicts that once the good economic cheer is fully dispensed, spending will increase by 4 percent. IDC's latest Worldwide Quarterly Server Tracker, released last week, reflected similar growth, with a 2 percent revenue pickup during the past three months. The gain is nearly one percentage point higher than analysts anticipated, and this was the second consecutive quarter of positive growth for server revenue worldwide.
Such optimistic growth indicators likely have some CIOs scrambling to expand their server-room wish lists, while others may be cautiously hoping they asked nicely enough for minimal add-ons. However, no matter how financially flush (or impoverished) you feel, now that it's late in the fiscal year and the creatures aren't stirring much, it's high time to assess the battle plan for next year and figure out how to get the most out of that budget now sitting in your inbox.
No matter how big the budget, a careful TCO analysis is in order. Calculations will help to realistically weigh server room price tags. Blade servers, for example, cost more up front, but they conserve space, cable clutter, and electricity, and they are beginning to prove themselves in enterprise environments. Hewlett-Packard is emerging as the market leader in this space, and one could argue that IA-64-based blades are moving mainframes closer to the retirement block. The migration to Linux-based servers evidenced in science, academia, and, modestly, in the enterprise may be initially costly, but open source has begun to prove its cost effectiveness. Examples of successful migrations to open source include Israel's Ministry of Commerce's OpenOffice 1.1 deployment and the Chinese government's and one Dutch company's commitments to Linux-based Windows alternatives, all of which have Microsoft taking (at times legal) notice.
CIOs who have determined (or for whom it has been determined) they can make do with a smaller budget than in the past also face a variety of choices. Grid and virtual computing, which developed during the past few years of cost-consciousness, continue to maximize server output as budgets modestly expand.
For a CIO looking to quickly pare down or contain costs, blades are generally not the way to go, as they will likely lead to purchasing more equipment than necessary. Blades aren't yet interoperable, and the vendors dominating the game, like HP and IBM, have little incentive to develop standards that would allow their competitors to sell low-cost alternatives. With standards nowhere in sight and prices unlikely to drop, purchasing a blade system may be less desirable than maximizing current server hardware usage through virtualization or a rudimentary grid.
The combination of standards (or lack thereof) and the hardware hodgepodge in many server rooms will likely result in interoperability remaining a key focus in 2004. A variety of operating systems have made interoperability easier in heterogeneous server rooms. For example, Panther's improved compatibility with Windows allows folder, printer, and network sharing between Mac and Windows users. Linux's adherence to open standards, coupled with enthusiastic backing from most major vendors, will enable the open source operating system to earn wider acceptance. As budgets increase, it will be interesting to watch whether enterprise interest in Linux, stoked partially out of a need to cut costs, will burn as bright.
It's important to protect even a slightly increased investment, and we recommend taking stock of potential server marginalization at this time. Whether bought or leased, server hardware should always provide a clear upgrade path as it approaches end of life. In this respect, ubiquity offers some guarantee of return on investment, as it outlines a frequently traveled upgrade path. A major vendor has the muscle to provide clearer retirement strategies for its servers than does a typical white box vendor. However, Windows, Linux, and other stripes of Unix can make for easier upgrade transitions than proprietary operating systems historically deployed by these bigger vendors.
Unfortunately, even the best laid plans are meaningless if an errant meteor razes the server room or some other unforeseen disaster occurs. Results of a recent Forrester Group survey found 62 percent of CIOs consider security or disaster recovery upgrades a top priority for 2004 IT spending. Administrators who neglect physical security can lose carefully tuned hardware to common theft as well as common environmental disturbances (e.g., blackouts or heat waves). No matter how tight the budget, security and failover should always be taken into account.