Benign ATL obsession: green computing

There's considerable interest in certain quarters in green computing--computing that's good (or at least better) for the environment. Accenture Technology Labs (ATL) shares this interest. Are we going all soft and altruistic? Well, yes (we're as concerned about the environment as the next research lab), but that's not the whole story. It turns out that green computing has significant hard benefits--over and above any good press you might receive from embracing it.

So What?

From an altruistic perspective, green computing is mostly about carbon footprints. One medium-size data center can generate 10,000 metric tons of carbon per year (because of the power it uses). Large carbon footprints are not exactly bad for business--not directly--but they can result in negative publicity. In that sense, going green can have PR benefits--you can position yourself as a socially responsible enterprise, at least with respect to IT.

On the hard benefits side, the key insight is that power is expensive and getting more so--to the point where Gartner VP Ramesh Kumar predicts that it could go from 10 percent of an IT budget to 50 percent "within a matter of years." Further, there are cities (and there will be more) to which power companies literally cannot deliver more electricity: Their grids are maxed out. According to ComputerWorld (March 26th, 2007), for example, "You can't buy any more power in Boston or Houston, and other cities are either on the tapped-out list or about to be. It doesn't matter if you are Warren Buffet or Bill Gates, you can't buy any more."

Clearly data centers need to become more energy efficient. And there are several approaches, some more exciting than others.

To begin with the mundane: air flow. For every watt consumed by a computer, another watt is required to cool it off. That number can be reduced as much as 30 percent by optimizing air flow in the data center, a significant saving.

More exciting (to a certain type of person): virtualization. The idea here is to take the applications that run on your servers (PCs) and cram them together in a smaller number of servers. Most servers are significantly under-utilized, which means there's plenty of excess "space" into which to pack the applications. Fewer servers = less power = green(er).

Of course, if you continue to grow, you're going to run into limits no matter how much air you redirect or how many applications you compress. Ultimately, the grid will fail you or power costs will be too high or you simply won't be able to pump enough cool air through the room. At that point, you need to take a deep breath and consider "utility computing." Under a utility computing model, an enormous central collection of servers is made available to enterprises for use as and when needed. It's unnecessary under this regime for an enterprise to own (or lease) its own servers--it just buys exactly as much computing power, disk space and bandwidth as it needs from moment to moment--and no more. As a result, the enterprise in effect never has any idle servers. In other words, in that sense, it's wasting no power at all.

Ah, nice try, you say. But what about the utility? Doesn't it have idle servers? Isn't it wasting power? Well, yes. But because it may have thousands of customers, it can share its excess capacity (available to handle spikes in demand) across all of them, which means it needs less excess capacity than its customers would if they were all operating their own independent data centers. So the utility model does indeed save electricity. A provider could even (and this is similar to what power companies do today) make arrangements with other utility computing companies to supply resources in times of unexpectedly high demand. You could also imagine a spot market in computing power.

Another aspect of utility computing: Google has a huge data center next to a hydroelectric dam in The Dalles, Oregon--and pays dramatically less for power than it would if it bought off the grid. (Microsoft and Yahoo are planning similar efforts in the same vicinity.) Industrial-scale computing facilities can be located so as to optimize across land costs, power costs, bandwidth costs and power reliability (Google's stated reason for locating near a dam). This can result in significant savings and is not an option most enterprises have available to them.

So there you have it: How to save money and be green at the same time. How often does life hand you an opportunity like that? Many of us here at the Labs are consumed by green computing and see it as a major trend for the next few years. To learn more, get in touch with Accenture Technology Labs' Teresa Tung (teresa.tung@accenture.com), who is leading the charge in this space.