On Cooperation Versus Competition Between Autonomous Resource Management Agents

Abstract

A common issue in distributed systems is how to optimize resource sharing when several resource management agents have access to the same resource pool. When the total resource demand reaches max capacity of the common pool, some strategy for resource sharing must be used. We compare “altruistic” behavior in which agents give up resources according to available evidence with “selfish” approaches in which agents with priority “steal” resources from others. Through simulation, we find that “altruistic” approaches provide closer to optimum behavior than “selfish” approaches, which instead lead to instability.