Return to Topics Reports 
General Information About The Author:

Name:            John Thrall
School:          Washington & Lee University
Address: 	 P.O. Box 4463
                 Lexington, VA 24450

Comments:    	 jthrall@wlu.edu

Topic Report: Grand Challenge Problems

The High Performance Computing and Communication (HPCC) defines a grand challenge as "a fundamental problem in science and engineering with broad economic and scientific impact, whose solutions can be advanced by applying high performance computing techniques and resources." These grand challenges include weather prediction, analysis of fuel combustion, ocean modeling, modeling the universe, etc.

A common theme in parallel computing, and in grand challenges in particular, is the use of the computer simulation paradigm. Computer simulation involves the use of a mathematical model to simulate a real world situation or problem, and then using computers to calculate the results of these mathematical models. Often the use of a simulation has many characteristics that make it much more desirable than actually performing an experiment in the real world. For example, a computer is now be constructed to perform computer simulations of nuclear weapons explosions. Obviously the ability to avoid actually testing the nuclear weapons provides considerable savings in terms of environmental as well as monetary costs.

Many computer simulation problems are inherently parallel, especially the grand challenge problems. Our goal is to provide accurate solutions to these computer models in a reasonable period of time. A reasonable period of time can range from several days in certain models to several hours if weather prediction is the goal. Obviously the weather prediction is no good if it is available only after the day it is predicting is already over!

In addition to computer modeling, a new set of grand challenge problems has arisen in the field of real time simulation. These involve the need to instantaneously perform a large number of calculations in order to provide real time data to researchers. For example, astronomers want radio telescope images to appear in real time so that they can use these images to point the instruments more accurately. Another application is to use virtual modeling techniques to allow more accurate controls of scanning electron microscopes.

To solve these grand challenge problems requires performance in the teraflop range (trillion floating point operations per second), and very large (100 gigabyte) memories. We have nearly reached our limit as to how fast a single proccessor can run, due to physical restrictions of the silicon used in the manufacture of computer chips as well as the speed of light. Faster speeds are being obtained using massively parallel machines as well as using very fast networks of machines.

The United States government issues grants yearly to new supercomputing projects across the country. Grand Challenge Awards represent some of the most current work in the grand chellenge feild of computing. The Grand Challenge Allocations Committee selects specific projects for NSF and other government funding. Some of the current research involves: