AbstractsComputer Science

Computing systems' complexity appears to be approaching the limits of human capabilities. To cope with the rapidly growing complexity of operating, managing and integrating computing systems, one of the most promising options is Autonomic Systems, which are computing systems that can manage themselves given high-level objectives from administrators. Self-management means that the system self configures at run time to increase responsiveness and agility, self-heals to improve business resiliency, self-optimizes to improve operational efficiency and self-protects itself from malicious attacks. While traditional approaches to computer systems are often centralized and hierarchical, Autonomic Systems are highly distributed with complex connectivity and interactions, rendering centralized management schemes infeasible. In this thesis, we describe the design and implementation of a decentralized architecture based on mobile multi-agent systems-an approach to building Autonomic Systems. Based on the proposed architecture, we developed a prototype application-an electronic marketplace, which achieves a set of desired features of Autonomic Systems such as autonomy, hiding complexity and self-healing. We put forth Autonomic Systems, a self-managing distributed computing system, as a new and promising application domain for multi-agent system ideas and argue that an Agent-based approach is well suited to construct Autonomic Systems.