Carnegie Mellon University researchers have successfully used a cluster of the same low-power processors that are used in netbooks and mobile devices to create a server architecture that uses less power than a light bulb. The architecture, called a fast array of wimpy nodes (FAWN), could decrease by an order of magnitude the amount of power consumed by the computational infrastructures at companies such as Google, Microsoft, Amazon, eBay, and Facebook. FAWN's developers say the new architecture could significantly improve the capabilities of cloud computing while decreasing its environmental impact.
The U.S. Environmental Protection Agency projects that by 2011, data centers in the United States could use up to 100 billion kilowatt-hours of electricity, costing a total of $7.4 billion and emitting an estimated 59 million metric tons of carbon dioxide. FAWN reduces energy consumption through a combination of relatively slow processors and flash memory, which creates an architecture with a performance per watt of energy ratio that is 100 times better than traditional servers. Flash memory has significantly faster random access than disk-based storage, and FAWN's slower processors require less power.
FAWN is limited to problems that require random access to small bits of information, but this type of input/output-intensive task is what strains the existing infrastructure of Web-based companies. FAWN has a major advantage over traditional systems because it eliminates the memory wall, which is caused by a disparity between the speed of the central processing unit and the speed at which data can be retrieved from disk-based storage.
From Technology Review
View Full Article
Abstracts Copyright © 2009 Information Inc., Bethesda, Maryland, USA