Chapter 33: Power

Power is the ultimate resource constraint of a planetary scale computer. Everything — computation, storage, networking, and cooling — requires electricity. The amount of power a data center can deliver to its servers determines the maximum compute capacity of the facility.

Power delivery in a data center involves multiple levels of redundancy. Utility power from the electrical grid is the primary source. Uninterruptible power supplies (UPS) provide battery backup for the seconds or minutes it takes for diesel generators to start during a grid outage. Generators can power the facility for days or weeks, limited only by fuel supply.

Power Usage Effectiveness (PUE) measures the ratio of total data center power to the power consumed by computing equipment. A PUE of 2.0 means that for every watt used for computing, another watt is used for cooling, lighting, and other overhead. Modern data centers achieve PUEs of 1.1 to 1.2, meaning that the vast majority of power goes directly to computation. Techniques like hot/cold aisle containment, free-air cooling, and liquid cooling contribute to these improvements.

The environmental impact of data center power consumption is increasingly important. Major operators are investing in renewable energy sources and carbon offsets. Locating data centers near renewable energy sources — hydroelectric dams, wind farms, solar installations — is becoming a key factor in site selection.