Most modern computers consist of one, or at most a few, processors that attack complex problems by dividing them into a linear sequence of operations. Speed is obtained by developing faster and more sophisticated processors.

Massively-parallel systems are fundamentally different. Imagine a computer consisting of 10,000 simple computers arranged in a structure that permits easy exchange of information from one processor to another. Great speed is achieved by allowing these 10,000 computers to run in parallel with each other.

Such systems are particularly effective for problems like pattern recognition and weather modeling where the computer must infer structures from an enormous collection of individual pieces of data.