Confidence in ASCI Scientific Simulations 1 HICSS-32, Software Technology Track MiniTrack-8 Parallel and Distributed Simulation Confidence in ASCI Scientific Simulations

The U. S. Department of Energy's (DOE) Accelerated Strategic Computing Initiative (ASCI) program calls for the development of high end computing and advanced application simulations as one component of a program to eliminate reliance upon nuclear testing in the U.S. nuclear weapons program. This paper presents results from the ASCI program's examination of needs for focused validation and verification (V&V). These V&V activities will ensure that 100 TeraOP-scale ASCI simulation code development projects apply the appropriate means to achieve high confidence in the use of simulations for stockpile assessment and certification. We begin with an examination of the roles for model development and validation in the traditional scientific method. The traditional view is that the scientific method has two foundations, experimental and theoretical. While the traditional scientific method does not acknowledge the role for computing and simulation, this examination establishes a foundation for the extension of the traditional processes to include verification and scientific software development that results in the notional framework known as Sargent's Framework. This framework elucidates the relationships between the processes of scientific model development, computational model verification, and simulation validation. This paper presents a discussion of the methodologies and practices that the ASCI program will use to establish confidence in large-scale scientific simulations. While the effort for a focused program in V&V is just getting started, the ASCI program has been underway for a couple of years. We discuss some V&V activities and preliminary results from the ALEGRA simulation code that is under development for ASCI. The breadth of physical phenomena and the advanced computational algorithms that are employed by ALEGRA make it a subject for V&V that should typify what is required for many ASCI simulations.