Utility-Theoretic Heuristics for Intelligent Adaptive Routing in Large Communcation Networks

Utility theory o ers an elegant and powerful theoretical framework for design and analysis of autonomous adaptive communication networks. Routing of messages in such networks presents a real-time instance of a multi-criterion quasi-optimization problem in a dynamic and uncertain environment. In this paper, we examine several heuristic decision functions that can be used to guide messages along a nearoptimal (e.g., minimum delay) path in a large network. We present an analysis of properties of such heuristics under a set of simplifying assumptions about the network topology and load dynamics. In particular, we identify the conditions under which one such utilitytheoretic heuristic function is guaranteed to route messages along an optimal (minimum delay) path in a network with uniform load except for a single hotspot (when it is assumed that the network load stays ArminMikler is currently a post-doctoral fellow at the Scalable Computing Laboratory of the Ames Laboratory (U.S. Department of Energy) at Iowa State University, Ames, IA 50011 Vasant Honavar is grateful to the National Science Foundation for its support of his research through the grant NSF IRI-9409580. The authors wish to thank Yong Zhao, Karthik Balakrishnan, Rajesh Parekh, and Chun-Hsien Chen (all doctoral students in Computer Science at Iowa State University), and Mary Oman (a post-doctoral fellow at Ames Laboratory) for their helpful comments on earlier drafts of the paper.