Multi-criteria decision-making for heterogeneous multiprocessor scheduling

Abstract

This paper proposes a new tri-objective scheduling algorithm called Heterogeneous Reliability- Driven Energy-E±cient Duplication-based (HRDEED) algorithm for heterogeneous multiprocessors. The goal of the algorithm is to minimize the makespan (schedule length) and energy consumption, while maximizing the reliability of the generated schedule. Duplication has been employed in order to minimize the makespan. There is a strong interest among researchers to obtain high-performance schedules that consume less energy. To address this issue, the proposed algorithm incorporates energy consumption as an objective. Moreover, in order to deal with processor and link failures, a system reliability model is proposed. The three objectives, i.e., minimizing the makespan and energy, while maximizing the reliability, have been met by employing a method called Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS). TOPSIS is a popular Multi-Criteria Decision-Making (MCDM) technique that has been employed to rank the generated Pareto optimal schedules. Simulation results demonstrate the capability of the proposed algorithm in generating short, energy-e±cient and reliable schedules. Based on simulation results, we observe that HRDEED algorithm demonstrates an improvement in both the energy consumption and reliability, with a reduced makespan. Speci¯cally, it has been shown that the energy consumption can be reduced by 5–47%, and reliability can be improved by 1–5% with a 1–3% increase in makespan.