I saw a headline the other day stating that, “Efficient Data Centers Increasingly Crucial for Telecom Companies”. After asking myself the obvious questions, “why just telecom companies?” and “does anyone think that they need an inefficient data center?”; it made me think about how flippantly the “E” word is tossed around today. If you’re like me, you’ve probably lost count of the articles that you’ve read, or conferences you’ve attended, where the big climatic summary has been that data centers must become more efficient—a comment greeted with a nodding of heads that makes the room look like it’s filled with bobble head dolls. Despite our universal agreement that we all think efficiency is good, does this mean that we are all talking about the same thing? It would seem that the best way to ensure the efficiency of your data center is to define exactly what you are attempting to achieve, and, in many instances, how much you’re willing to pay for it.
The vast majority of data centers being built today are inherently inefficient on one or more levels. For example, if you feel that you data center needs will grow over time from 10,000 square feet to 50,000, is building the entire shell out first efficient? If you’re concerned about the allocation of your capital, the answer would be no. Spending $20-$100 million for a building that you may never fill—sometimes reality has a way of contradicting our best estimates—is probably not the best use of your company’s money.
Building the entire shell upfront is terribly inefficient when you consider the other potential uses for the land it sits on. If you build you data center of the present and the future all at once, you’ve tied that land up for the next 20 years or so. More succinctly, you’ve eliminated your ability to use that land for any other purpose, including building some other corporate asset or even selling it. From an efficiency standpoint this mode of implementation renders an appreciating asset (land) distinctly inefficient.
Since most data centers are built using non-standard designs (even those that claim that they aren’t), the amount of materials used in the construction of each of them vary dramatically. If examined strictly from a construction perspective this situation would seem to have impact on efficiency since the amount of material needed to build the facility is just that, the material necessary to build the specified site. However, when the CO2 generated for the production of just a cubic yard of pure concrete staining (400 lbs.) or one (1) ton of steel (1.9 tons) is factored into the equation, the efficiency of the site’s design and materials usage must be re-examined.
Even in areas seemingly as straightforward as PUE, the issue of the efficiency of the methodology used to drive this metric lower must be weighed against its costs. If an evaporative cooling system is your choice to lower your site’s PUE another tenth of a point, you must factor in the on-going cost of millions of gallons of treated water the system will require annually along with your initial system cost in evaluating the practicality of your decision.
No one disagrees that data centers should be “efficient”. Unfortunately, it seems that the definitional scope of the term is frequently too narrow to accurately reflect all the areas of data center design, construction and operation that fall within its purview. In making their data center decisions, today’s customers should be mindful that it is their definition of the term, and the cost they are willing to pay to achieve it that must drive their efforts; not the nebulous pronouncements from industry pundits and “experts” who should know better.