Task Allocation in Heterogeneous Computing Environment by Genetic Algorithm

Abstract

This paper presents an efficient technique for mapping a set of tasks onto a set of heterogeneous processors. The tasks require data communication between them. The system is assumed to be completely heterogeneous, where the processing speeds, memory access speeds, communication latency between processors and the network topology are all considered being non-uniform. Typically, the numbers of tasks are much larger than the number of processor available. The problem of optimal mapping of the tasks to the processors such that the application run -time is minimized is NP-Complete. The searching capabilities of genetics algorithms are utilized to perform the optimal/near optimal mapping.