In an article at Network World, Patrick Thibodeau discusses a National Resources Defense Council (NRDC) report that takes the data center industry to task for contributing to pollution by wasting energy. Perhaps the most illustrative statement in Thibodeau’s piece is the opening line: “U.S. data centers are using more electricity than they need.” Of course, wasting resources is undesirable, but at what point do conservation efforts become counterproductive?
Taking Conservation to the Extreme
It’s probably safe to say that, with very few exceptions (at least in the developed world), all consumers and businesses are using more electricity—and other resources—than they need for survival. Think of typical consumers: chances are, they are less than zealous about turning off lights in unoccupied rooms, using appliances in an efficient manner, avoiding electricity “vampires,” keeping the thermostat set at livable (not necessarily comfortable) temperatures and so on. Companies likewise may trade some efficiency for some other business good, like lower capital or labor expenses, convenience and so on.
The point here is simply that there is always room for greater efficiency. The question is what should drive efficiency increases, and who is to say when enough efficiency is enough?
Economics as the Driving Force
Of course, all bets are off for a population that discards all restraint and morality. But assuming there is at least something of a collective conscience, the question of how much waste is too much is a matter of economics. Again, consider the case of a consumer: although leaving every light in the house on all the time may be convenient (you can always see where you’re going), it can get expensive. Most consumers therefore feel some need to conserve energy simply because their income is limited and the conveniences of leaving the lights on fail to outweigh a painful power bill. But that doesn’t mean that all unoccupied rooms should always be immersed in darkness—sometimes convenience (or safety, or some other consideration) is worth the extra cost.
Similarly in the case of data centers, some companies may, for instance, view the ongoing operational costs of less efficient servers to be worth saving the capital costs of replacing them with newer, more efficient models. Is this the best choice? Who knows, unless all considerations are weighed and compared via some sort of objective calculus (something that is no simple matter to develop).
But there is a check on a company’s wastefulness: economics. In one case, a company might hemorrhage too much money in excess energy costs to sustain itself, ultimately going out of business as a consequence of its wastefulness. In another case, its lack of environmental consciousness may drive consumers to “greener” competitors, forcing it out of business for a lack of income. Today’s data center industry—although it is a growing consumer of energy, and not a very efficient one at that—has yet to run up against this constraint. That isn’t to say there’s no room for meaningful, beneficial and even profitable improvement, but the answer must go beyond simply harping on companies whose PUEs don’t meet some magical value chosen arbitrarily by—well, someone.
Energy efficiency is not a zero-sum game, at least if the so-called Jevons paradox has any basis in reality. This paradox proposes that as energy efficiency increases, the rate of energy consumption tends to increase as well. In other words, higher efficiency doesn’t always lead to decreases in usage across an economy or industry. For instance, looking at the small scale, a company that installs power-efficient servers might see the energy savings not as money to stuff in its pockets but as opportunity for expansion of the business, consuming some or all of those savings. Of course, individual cases may go either way, but the observation underlying the Jevons paradox is that when such efficiency increases are applied broadly, the tendency is toward greater consumption rather than less.
Some economists naturally dispute this claim, but it does make a certain amount of intuitive sense. Assuming it is true even to a small extent, then, efficiency increases—as beneficial as they are (broadly applied, they can increase the standard of living at a given level of energy consumption)—can’t be the sole solution to the problem of growing energy consumption by data centers (or the economy as a whole).
What to Do About Data Center Efficiency
Journalism and studies examining the efficiency of data centers do serve a useful role in that they can make companies and consumers aware of the issue, potentially changing consumer behavior as well as business decisions. The NRDC report offers several recommendations, including use of a “simple server utilization metric” and public disclosure of data center efficiency. As voluntary actions, these practices could be beneficial, although some companies will no doubt find ways to “game the system”—just as they do with PUE, for example.
Since demand for a resource falls as price rises, one fair option is an end to government energy subsidies. Subsidies effectively reduce the cost of energy production, and assuming at least some of those savings are passed to consumers (businesses and individuals), the demand for energy rises. Energy at unsubsidized market prices will be consumed at a lower rate, leading to less pollution.
Another approach to efficiency is to start at the “grass roots”: inform consumers and encourage them to adopt energy-efficient practices first. A movement tends to be stronger when it moves from a broad base up rather than a narrow peak down; companies stay in business by meeting consumer demand—regardless of their social-responsibility policies and so on. Citing the NRDC report, Thibodeau notes that “IT managers are ‘extremely cautious,’ about implementing aggressive energy management because it could introduce more risk to uptime.” This sentiment is unsurprising: the average consumer, even a “green” one, is more likely to experience a temper tantrum over a downed service than a little too much energy consumption by the service provider.
Government mandates for efficiency can also help the situation, at least on the surface, but they may cause more harm than good. Furthermore, government agencies may be the worst culprits: the bureaucratic mentality requires than whenever an agency garners some savings on a budgeted item (like, say, energy consumption), those savings must be consumed lest next year’s budget for that item is cut—the Jevons paradox at its worst. Furthermore, the unsustainability of almost everything the government does (from its expanding debt to its unending foreign military escapades) largely disqualifies it from moral authority in the area of conservation. It’s up to companies and consumers to do what’s right.
The data center industry isn’t as efficient as it could be—and neither is any other industry, company or individual. Efficiency is often sacrificed for the sake of convenience or even survival, and that’s not necessarily wrong. To be sure, however, wasting energy can be doubly destructive, in that energy production has serious environmental costs, but so does every activity. Building a house might, for instance, involve reducing available habitat for certain animals (but it might increase available habitat for others). As with any environmental issue, the key is balance: recognizing that every activity (or lack thereof) has upsides and downsides, and making choices in a manner that provides the most benefits at the lowest economic, social and environmental costs. Efficiency is beneficial, but berating a lack of efficiency without recognizing other issues is shortsighted.