Distributed Computing and Markets

It used to be that access to a single computer processor was a valuable commodity. Computer time was doled out in little dribbles measured in thousandths of second. Computer processing power was so valuable that it was only devoted to solving important problems of national scope.

Today, screen savers which paint pretty moving designs on computer screens to extend the life of monitors consume enormous quantities of computer time. The processing power devoted to drawing fanciful geometric figures on screens represents an enormous and largely untapped computing resource. In this month's American Scientist, Brian Hayes details how the cheapness and ubiquity of computing power is leading to a new paradigm for solving seemingly intractable problems.

Hayes reports that the world's fastest computer named Janus is owned by Sandia National Laboratory and is used to simulate nuclear explosions. This massively parallel computer is powered by 9,216 Pentium Pro processors. By anyone's standards, that is a lot of processors, but there are probably large retail stores in suburban America that have sold that many processors. Even Janus's computing power is minuscule compared to the total computing power connected through the Internet.

There have been a number of attempts to harness this power and solve problems that even Janus would choke on. A prime number is a number evenly divisible only by 1 and itself. The measure of supercomputing power used to be how large a prime number a computer could produce. Today the records for the largest prime belong to groups of individual computer owners who share computation across the Internet.

One commercial application of this approach to problem solving is testing encryption schemes. RSA Data Security regularly checks the robustness of its encryption algorithms by offering monetary rewards to those who can decode publicly available messages. In one such contest, Hayes reports that a group had organized enough computers and their owners that encryption keys were being tested at a ``rate equivalent to the work of 26,000 Pentium computers,'' or three times faster than Sandia's Janus.

The movement of large computing problems from large central facilities to distributed processing by ordinary machines is a metaphor for the operation of the economic markets in allocation of economic resources. No matter how large a bureaucracy and its clever clerical cadre, it is difficult to match the wisdom of the combined intelligence of millions of ordinary minds. Think of a market as a large distributed computer. Instead of Internet packets, information is exchanged through monetary exchanges, in an effort to find optimum solutions to resource allocation problems.

Go To Page: 1 2

Articles in this Topic    Discussions in this Topic