Grid Enabled Distributed Computing: from Molecular Dynamics to Multiscale Simulations

Abstract

Molecular Dynamics is an approach suited to face several computational grand challenges which involve the determination of the behaviour of chemical system in energy, environment and technology applications. As such molecular dynamics prompts huge requests of computing resources difficult to match even using the supercomputers available at large scale computing facilities. At present, a promising alternative to the not always accessible supercomputers available at big supercomputer centers is distributed computing on Grid platforms. The possibility of exploiting the Grid technologies and in particular the resources made available by the European Grid project has allowed us to carry out massive computational campaigns after implementing our programs on the EGEE Grid production infrastructure. These efforts have materialized into the constitution of a Virtual Organization (VO) called COMPCHEM whose main goal is to gather together computational researchers willing to calculate the key properties of molecular systems in an ab initio fashion. This is indeed the approach adopted in the present thesis with the study of three families of computational chemistry applications.</p><p>Accordingly the thesis is articulated as follows. The first section deals with the theoretical aspects by starting from first principle equations of quantum reactive scattering to land into the classical mechanics approach. The first section continues by discussing the problem of representing molecular interactions and statistically averaging detailed information in order to evaluate observable properties and use them as molecular engines of the microscopic level of multiscale simulations. The second section deals, instead, with the discussion of the evolution of the ICT technologies from parallel to various forms of concurrency. Then among concurrent technologies the modern distributed computing platforms are discussed in detail. In the same section the innovative features of the production Grid of EGEE and of the Virtual Organizations built on top of it, are discussed. The third section deals instead with some computationally demanding molecular science applications.