Electronic Thesis and Dissertation Repository


Doctor of Philosophy


Computer Science


Michael A. Bauer


The complexity of today’s data centers has led researchers to investigate ways in which autonomic methods can be used for data center management. Autonomic managers try to monitor and manage resources to ensure that the components they manage are self-configuring, self-optimizing, self-healing and self-protecting (so called “self-*” properties). In this research, we consider autonomic management systems for data centers with a particular focus on making data centers more energy-aware. In particular, we consider a policy based, multi-manager autonomic management systems for energy aware data centers. Our focus is on defining the foundations – the core concepts, entities, relationships and algorithms - for autonomic management systems capable of supporting a range of management configurations. Central to our approach is the notion of a “topology” of autonomic managers that when instantiated can support a range of different configurations of autonomic managers and communication among them. The notion of “policy” is broadened to enable some autonomic managers to have more direct control over the behavior of other managers through changes in policies. The ultimate goal is to create a management framework that would allow the data center administrator to a) define managed objects, their corresponding managers, management system topology, and policies to meet their operation needs and b) rely on the management system to maintain itself automatically. A data center simulator that computes its energy consumption (computing and cooling) at any given time is implemented to evaluate the impact of different management scenarios. The management system is evaluated with different management scenarios in our simulated data center.