Parallel Computing Using Linux
Various classes of problems lend themselves to parallel computing solutions. This article discusses the concepts and shows how Linux can be used to address the problem.
Parallel computing involves the design of a computing system that uses more than one processor to solve a single problem. For example, if two arrays with ten elements each must be added, two processors can be used to compute the results. One processor computes the sum of the first five elements and the second processor computes the sum of the second five elements. After the computation, the results from one processor must be communicated to the other processor. Before starting the computation, both processors agree to work on independent sub-problems. Each processor works on a sub-problem and communicates when the solution is available. Theoretically, a two-processor computer should add the array of numbers twice as fast as a single-processor computer. In practice, there is overhead and the benefits of using more processors decrease for larger processor configurations.
Affordable Supercomputers
Obtaining a Unix workstation for the cost of a PC has been one of the benefits of using Linux. This idea has been carried a step further by linking together a number of Linux PCs. Several research projects are underway to link PCs using high performance networks. High speed networking is a hot topic and there are a number of projects using Linux to develop a low latency and high bandwidth parallel machine. (One URL is http://yara.ecn.purdue.edu/~pplinux.)
Read more...
http://www.linuxjournal.com/article/1350 Don
No comments:
Post a Comment