U.S.-France (INRIA) Cooperative Research: Parallel Numeri- cal Methods for Transient Solutions of Markov Chains

点击翻译按钮获取中文摘要

This three-year award supports U.S.-France (INRIA) cooperative research in parallel computational methods in Markov Chains. William Stewart of North Carolina State University and Bernard Phillipe of INRIA in Rennes, France are the U.S. and French coordinators. The project involves additional U.S. and French researchers in parallel computing, computer performance evaluation, parallel algorithms and numerical methods. The investigators propose to examine new methods for finding equilibrium solutions of large, sparse Markov Chains using parallel computers. They propose new techniques for computing special cases of stochastic matrices and small matrices and to examine a variety of computer architectures using vector processors, shared memory multiprocessors, distributed memory multiprocessors and massively parallel architectures. The project will benefit from complementary expertise and sharing of essential equipment. The U.S. and French investigators bring to this collaboration expertise in computer science and applied mathematics. The essential equipment for doing parallel processing architectures are available at participating U.S. institutions and at INRIA, the French national research institute for information science and automation. INRIA is a world class institute in computer and information science and leads a European consortium of research organizations in these fields. Markov Chains are useful mathematical devices for describing real systems and queuing networks. The proposed research will be very important in telecommunications, computer performance and manufacturing systems modeling. Advances in this area will have far-reaching effects since Markov Chains are models used in other areas of science and engineering.