08/01/1994 09:00 AM
Computer Engineering
As parallel systems become larger and more complex, they become harder to understand. The vast amount of information produced by these systems can quickly overwhelm a person. In many cases, researchers have turned to the use of visualization techniques as an aid to understanding the inner workings of parallel systems. By converting the information to images, a visualization system can take advantage of the human ability to recognize patterns that would not seem obvious otherwise. There are several existing systems for visualizing parallel programs. But most of them present fixed views of data, and will not scale to the thousands of processors available on massively parallel computers. In this paper we present a proposal for research in developing visualizations that scale with the number of processors. The methods used to create these views will make visualization systems useful in understanding computers with any number of processors.