META Report: Room at the Data Center?Rich Evans
Driven by centralization, ongoing growth, and sustained expansion of both storage and processor infrastructure, businesses are re-examining their short-term data center growth plans.
More important, while base processor and storage form factor technology will see densities double every 18 months, adding complicated power supply and expanding interconnection requirements will throttle actual floor space improvements. On average, only about 14 percent to 17 percent of floor space will be freed up each year.
|Other Recent META Reports|
|Java Keeps Brewing, But .Net Looms The Hazards Of Subscription Licensing Doors Still Closed To Open Source Databases|
Dramatic growth in capacity, especially for storage, will make significant demands on data center real estate and power. As significant as pure floor space, power upgrades will be an ongoing event, requiring a doubling in power density every three to five years. For example, a relatively modest 30 percent compound annual growth rate in hardware capacity can increase real floor space requirements by 40 percent to 50 percent in three years (an annual rate of about 13 percent).
More typically, 50 percent annual capacity growth rates will force data centers to more than double during the same period (a 30 percent annual growth rate). With data center floor space costing $700-$1,000 per square foot for first-class accommodations, a rapidly growing 10,000-square-foot data center will face additional costs of $8 million to $12 million over a three-year period.
To avoid prolonged budget and possible infrastructure failures because of poor implementations, users must develop strong, realistic data center growth planning disciplines that meld technology refresh cycles, infrastructure costs, and physical plant expenses into a realistic game plan.
Technology Space Folk Law
An ongoing debate among physical planners is the battle around new technology versus old, and the effects on floor space. Although important, we believe the debate is a bit narrow. Moreover a "best" space solution (driven almost always by newer technology) must also include additional technology costs and physical plant upgrades, such as power and air conditioning.
The pros and cons of the argument can quickly be resolved with the help of a model that measures the effects of three- versus four-year technology refresh cycles. For example, for the modest-growth shop (30 percent annual capacity growth), space differences between a three- and a four-year refresh plan can vary between 5 percent and 11 percent in space because of infrastructure product cycles. With a 10,000-square-foot data center as a baseline and a three-year replacement cycle, this center will grow to 14,260 square feet in three years; a four-year fresh cycle results in the need for an additional 1,570 square feet, or $1.57 million in additional floor costs.
The driver of this difference is that even though a large (percentage wise) technology refresh occurs each year on average -- 42 percent for the three-year model and 35 percent for four-year model -- older, larger technology makes up a larger portion of the installed base for longer periods. So even though a four-year refresh model requires about 17 percent less technology and users might be able to delay one year an inevitable power and air upgrade, simple space costs will drive a significant a portion of technology refresh cycle decision making.
For companies planning on adding technology capacity by 50 percent a year, refreshment cycles have less effect on floor space. For example, a three- versus four-year refreshment cycle can result in a 4 percent to 10 percent floor space increase. Again, looking at our 10,000-square-foot example, the floor space needed after three years is 21,970 square feet. A 10 percent increase puts needed floor space at 24,167 square feet, and incremental costs for this additional 2,197 square feet could be as high as $2.2 million.
For the four-year technology refresh model, because the company is buying less technology for every period, equipment savings produce only 12 percent technology savings. For many users, the additional space costs point to shorter technology cycles as an overall better solution. (They do, however, require substantial planning, as well as often grueling capital appropriation cycles.)
Other, more-aggressive growth models (e.g., 70 percent or even 100 percent) produce similar results: newer and shorter technology life cycles help win the space war, with fewer and fewer differences in space requirements for three- and four-year refresh cycles. However, in all cases it is worth modeling the technology landscape for at least five years to determine short-term advantages driven by complicated power density requirements.
Nasty Power Density
Many traditional data centers have been undergoing significant infrastructure upgrades or replacements required to support rapid storage and rack-mounted server growth, which usually happens in conjunction with consolidation. However, many users are becoming painfully aware of the emerging environmental deficiencies (e.g., power, cooling, distribution, connectivity) of near-term facility planning, as well as their ominous threat to long-term growth needs. For example, late-1990s' requirements of 100 watts per square foot of average power now exceed 200 watts per square foot, and will double again by 2004.
Driven primarily by rapidly increasing rack-mounted servers (being added at 45 percent to 55 percent annual rates) and storage, power-hungry hardware threatens many current data-center design specifications. Although a slower technology refresh cycle (at the expense of space) can push power upgrades out 12-18 months, these power and cooling problems will continue to plague designs.
In fact, the 2004 400-watts-per-square-foot requirement will again double by 2008, so designers must work this into longer-range facilities plans so that space can be fully used, as opposed to 40 percent to 50 percent utilization because of insufficient power and cooling.
Business Impact: Insufficient investment in data center infrastructure can minimize customer transaction, interactions, and sales volumes, which can negatively affect overall revenue.
Bottom Line: To avoid the risk of infrastructure shortfalls caused by insufficient facilities size and functionality, users must implement strong data center growth planning disciplines that include model technology refresh alternatives as well as power density requirements.
Rich Evans is a consultant with META Group, an IT consulting firm based in Stamford, Conn.