HR: 10:20h
AN: SF32A-01 INVITED     [Abstracts]
TI: A Distributed Computing Infrastructure for Computational Thermodynamic Calculations of Solid-Liquid Phase Equilibria
AU: * Ghiorso, M S
EM: ghiorso@geosci.uchicago.edu
AF: University of Chicago, Department of Geophysical Sciences 5734 S Ellis Ave, Chicago, IL 60637 United States
AU: Kress, V C
EM: kress@u.washington.edu
AF: University of Washington, Department of Earth and Space Sciences, Box 351310, Seattle, WA 98195 United States
AB: Software tools like MELTS (Ghiorso and Sack, 1995, CMP 119:197) and its derivatives (Ghiorso et al., 2002, G3 3:10.1029/2001GC000217) are sophisticated calculators used by geoscientists to quantify the chemistry of melt production, transport and storage. These tools utilize computational thermodynamics to evaluate the equilibrium state of the system under specified external conditions by minimizing a suitably constructed thermodynamic potential. Like any thermodynamically based tool, the principal advantage in employing these techniques to model igneous processes is the intrinsic ability to couple the chemistry and energetics of the evolution of the system in a self consistent and rigorous formalism. Access to MELTS is normally accomplished via a standalone X11-based executable or as a Java-based web applet. The latter is a dedicated client-server application rooted at the University of Chicago. Our on-going objective is the development of a distributed computing infrastructure to provide "MELTS-like" computations on demand to remote network users by utilizing a language independent client-server protocol based on CORBA. The advantages of this model are numerous. First, the burden of implementing and executing MELTS computations is centralized with a software implementation optimized to a compute cluster dedicated for that purpose. Improvements and updates to MELTS software are handled locally on the server side without intervention of the user and the server-model lessens the burden of supporting the computational code on a variety of hardware and OS platforms. Second, the client hardware platform does not incur the computational cost of performing a MELTS simulation and the remote user can focus on the task of incorporating results into their model. Third, the client user can write software in a computer language of their choosing and procedural calls to the MELTS library can be executed transparently over the network as if a local language-compatible library of routines is being accessed. Fourth, the flexibility of calling library functions means that the client has more control over the configuration and output of the MELTS calculation. Fifth, if the client computer is a multi-processor compute cluster capable of issuing parallel requests to the MELTS "remote" library, then these requests may be in turn parallelized to the server compute cluster to enhance throughput and performance. Application of this computational model to fluid dynamical simulations of melting and transport in the Earth's mantle is envisioned. Further information and example clients for utilizing the current prototype library for distributed computing applications can be found at http://melts.uchicago.edu.
DE: 3600 MINERALOGY AND PETROLOGY (replaces
DE: 3630 Experimental mineralogy and petrology
DE: 3640 Igneous petrology
DE: 3694 Instruments and techniques
SC: Special Focus: Advances in Data Acquisition, Management, Analysis and Display [SF]
MN: 2004 AGU Fall Meeting