Texture-Based Heuristics for Scheduling Revisited

Abstract

Recent scheduling work has challenged the need for sophisticated heuristics such as those based on texture measurements. This paper examines these claims in the light of advances in scheduling technology. We compare a number of current heuristic commitment techniques against a texture-based heuristic. Our results demonstrate that texture-based heuristics can outperform these widely-used heuristic commitment techniques. Introduction Our research goal is to be able to model and quickly solve scheduling problems as they exist in the real world. We are less interested in optimal solutions than in fast approximate solutions: a quickly found solution that takes into account all the constraints in the real problem is of significantly more use than an optimal solution that either takes too long to find or does not accurately represent the problem. We are applying and extending constraint-directed scheduling techniques toward this end. Our search philosophy is to spend significant but ...