Multi-Objective Scheduling

Abstract

This chapter considers multi-objective scheduling, i.e., schedul- ing with simultaneous optimization of several objectives. The main motivation to optimize more than one objective is the growing complexity of modern sys- tems. Characterizing by one variable only the performance of a heterogeneous system with many users shows a narrow view of the system. In multi-objective scheduling, one can explicitly model, optimize, and find trade-offs between various performance measures. We introduce multi-objective scheduling on three motivating problems to illustrate different approaches. In MaxAndSum, a scheduler optimizes both the makespan and the sum of completion times. In EfficientReliable, the goal is to find the trade-off between the makespan and the reliability of schedule on failing processors. Finally, in TwoAgentMinSum, a processor must be shared fairly between jobs produced by two independent users. We study the complexity of these three problems and propose different approaches to obtain an acceptable solution. In MaxAndSum, an algorithm