The Parallel Coupling Problem
Dr. J. Walter Larson (ANUSF/APAC, Australian National University)
MSI Advanced Computation SeminarDATE: 2006-05-01
TIME: 11:00:00 - 12:00:00
LOCATION: GD 35 (John Dedman building)
CONTACT: JavaScript must be enabled to display this email address.
ABSTRACT:
Scientific simulation is progressing from the traditional approach of simulating individual processes and systems towards multiphysics simulation of complex systems comprising many mutually interacting subsystems. A classic example of this situation is a coupled climate model. Additionally, scientists are beginning to tackle simulation of multiscale systems that involve differing physics over multiple interacting length scales. The key technology enabling these bold enterprises is parallel computing, and in particular message-passing parallelism. This technological boon creates a new obstacle--the parallel coupling problem (PCP)--in the conveyance and transformation of data between multiple, distributed-memory applications. This presentation will begin with a brief survey multiphysics and multiscale applications, and how the computational complexity of their constituents give rise to the PCP. This will be followed by a discussion of the PCP and the challenges it presents in both coupled model architecture and parallel data processing. A brief overview of relevant available and emerging software technologies will be presented. Finally, a set of open problems will be stated.
BIO:
Dr. Larson began his career as a theoretical physicist, working in the areas of nonlinear wave theory, plasma physics, and dynamical systems, earning a PhD from the College of William and Mary. After his PhD he switched fields to climatology, and during this period of his career he became interested in the computational issues surrounding climate, weather, and ocean modelling applications. From 1996-1999 he worked at the NASA Data Assimilation Office in their analysis group on their operational data assimilation system. In 1999 he joined the Mathematics and Computer Science Division at Argonne National Laboratory, where he has worked for several years on the computational challenges presented by coupled climate modelling, and was the co-lead architect of both the Model Coupling Toolkit, and the parallel flux coupler (CPL6) that is in the National Center for Atmospheric Research Community Climate System Model. In 2003 he was appointed as a Senior Fellow in the Computation Institute at the University of Chicago. He has been active in the Common Component Architecture project since 2001, and is currently one of the leaders of its Parallel Coupling Infrastructure initiative. In February, 2006, Larson joined ANUSF, where he works at present.
