Laxmikant V. Kale, Klaus Schulten, Robert D. Skeel, Glenn Martyna, Mark
Tuckerman, James C. Phillips, Sameer Kumar, and Gengbin Zheng.
Biomolecular modeling using parallel supercomputers.
In S. Aluru, editor, Handbook of computational molecular
biology, pp. 34.1-34.43. Taylor and Francis, 2005.
KALE2005
Our knowledge of molecular biology and the machinery of life has been
increasing in leaps and bounds. To coalesce this knowledge into a
deeper understanding, we need to determine the structure of a
multitude of proteins with high resolution, and understand the
relationship between their structure and function. Molecular dynamics
simulations help further this understanding by allowing us to observe
dynamical phenomena occurring at an atomic level, and validate our
understanding of the basic physical principles embodied in
simulations. Simulations based on classical mechanics, with some
approximations of the quantum-mechanical ``reality'' are adequate for
many situations; however, for simulations involving making and
breaking of bonds, for example, a quantum mechanical simulation is
necessary. The Car-Parinello algorithm and the ability to combine
classical and quantum models in a single simulation are efficient ways
of accomplishing this.
In either case, the computational power needed for carrying out the
simulations over an interesting interval of time of the biomolecular
phenomena is so large that only parallel computers offer the hope of
completing such simulations in a realistic time. Although large
parallel computers are available now, it is quite challenging to
parallelize the simulations so as to scale to thousands of processors
and beyond. This paper presented an overview of strategies aimed at
this problem, and presented in some detail the particular strategies
the authors have been pursuing.
