Large-scale parallel and grid computing for turbulence and biomechanics
problems

Steven Dong, Division of Applied Mathematics, Brown University

Solution of grand-challenge problems in physical and biological sciences
such as direct numerical simulation (DNS) of drag crisis in bluff-body
flows and modeling blood flow in the entire human arterial tree requires
very-large-scale parallel and grid computing due to the enormous problem
sizes. The key question facing computational scientists is how to devise
efficient algorithms to exploit the power of supercomputing and the
essentially unlimited scalability of the computational grids such as the
TeraGrid. In this talk I will discuss multilevel-type parallel paradigms
within the spectral element framework that can take advantage of the
hierarchical structures arising from the physical problems and
computations as well as those in the architectures of supercomputers and
computational grids.  These algorithms have enabled simulations of
turbulence and biomechanics problems with hundreds of millions of degrees
of freedom. I will specifically look into the DNS of turbulent bluff-body
flows at Reynolds number Re=10000 and the simulation of blood flow in the
human arterial tree on the TeraGrid.