Learning-based Coordination of Distributed Component Deployment

Abstract

Self-organizing and resource-aware component deployment is an important feature of mobile pervasive systems. Distributed resources must be dynamically allocated to software components to ensure QoS demands and not distracting the user. In this paper, we propose a Reinforcement Learning technique to optimize distributed component deployment and migration. We argue that the approach meets some main requirements demanded by applications running on mobile systems. A motivating scenario is presented in which a distributed application server allows users to share content and run applications in mobile ad-hoc networks.