Data centers aren't energy efficent? Where did that come from?
This morning I ran across a couple of articles - a piece on the electricity usage of data centers by James Glanz of the New York Times; and a thorough rebuttal by Guy-Who-Knows-His-Stuff Diego Doval. Together, the articles got me worked up enough to write a post for the first time in nine months.
Glanz's article boils down to "data centers use a crazy amount of electricity," and implies that "data centers are indifferent, inefficient users of electricity." In response, Doval provides a step by step breakdown of nearly everything that is wrong in the NYT article. For a moment I want to use some rough math to reinforce one of Doval's points - that Glanz's accusation of the industry's indifference to energy efficiency is flatly wrong.
Worldwide, data centers accounted for about 0.5% of electricity use in 2000. That percentage grew to roughly 1.3% in 2010. As worldwide electricity consumption also increased from 2000 to 2009 (sorry, couldn't find 2010 numbers) by around 31%, we can calculate that absolute power usage by data centers worldwide grew in the neighborhood of 350% between 2000 and 2010. During that same period, the number of Internet users grew from 304 million to 1.97 billion. That is roughly a growth of 550% in users alone, and does not capture the significant per user growth in data transmission, storage, and processing that also occurred in that decade. In other words, 550% just starches the surface of growth in demand for computing, networking, and storage resources. But with 550% growth in demand as a floor, we can see the demand placed on worldwide computing resources has clearly grown faster than the energy consumption of those computing resources.
The above numbers represent something that is pretty well known: the computing industry has and continues to improve its energy efficiency. In fact, the computing and data center industries are relative utopias of energy efficiency compared to most other segments of the economy. What other segments have equipment turnover that is measured in months and consistently improves energy efficiency each generation? What other market segments have seen energy efficiency gains that roughly match demand growth? Compared to most other industries, computing and data storage are poster children for addressing energy efficiency.
What frustrates me about Glanz's article is that such a highly visible article is so off the mark. If you want to focus public attention on an energy efficiency issue, why focus attention on an industry that is largely addressing it's issue? How about an article on aging home appliances? Or aging automobiles? Or aging power plants?
Furthermore, the data center and computing industries are not perfect. Want to hold those industries accountable? How about an article on asinine patent wars? Or a continuing disregard for user privacy? Or a reliance on an ad based revenue model that may well collapse any day? Want to take those industries to task on environmental grounds? That high turnover I mentioned creates mountains of environmentally toxic e-waste every year. Why bend over backwards to frame an industry's strength as a deficiency when there are so many actual deficiencies to highlight?