NOTHING COSTS MORE THAN IT USED TO:

Computer scientists from Carnegie Mellon University have devised a framework for running large-scale computations for tasks such as social network or Web search analysis efficiently on a single personal computer.

The software could help developers working on many modern tasks: for example, designing a new recommendation engine using social network connections. In order to make effective recommendations--"your friends liked this movie, so here is another movie that you haven't seen yet, but you will probably like"--the software has to be able to analyze the connections between the members of a social network. This type of task is called graph computation, and it is increasingly common. But working with large-scale data sets (such as online social networks) usually requires the processing horsepower of many computers clustered together, such as those offered by Amazon's cloud-based EC2 service.

The new software, called GraphChi, exploits the capacious hard drives that are becoming ever more common in personal computers.

Follow @brothersjudd