Partitioning Problems in Heterogeneous Computer Systems

Abstract

Heterogeneous Computer Systems (HCS) have the potential to achieve enhanced performance and costeffectiveness over hLomogeneous systems for solving problems with diverse computational requirements. As the HCS consist of number of parallel computers which may have distinct language and system support and sometimes different computational models, the algorithm designer faces the formidable task of optimally assigning the subtasks of an application onto various machines of the HCS. In this paper, we address the problem of finding the least expensive set of super-speed machines subject to a maximal execution time constraint for a given application task. We show here that it is possible to find an optimal solution to the above problem under certain restrictions on the interconnection structure of the subtasks as well as the heterogeneous computer system. The algorithm that we have presented in this paper can also find a solution to the dual of the above problem i.e. finding the fastest set of machines to solve a given application task subject to a given cost constraint.