Adaptation Strategies in Policy-Driven Autonomic Management

Policies have been proposed as a means to express required or desired behavior of systems and applications. Policies can be used within the autonomic management elements of the system to provide action directives to adjust application or system tuning parameters in order to meet operational requirements. Policy specification is frequently component-based, that is, focused on the operational requirements of a particular system component, e.g., a Web server or a database. In multi-component systems, such as e-commerce systems, multiple components may co-exist on a single server and cooperate to deliver a set of services. It is reasonable to expect that each component would have its own set of associated policies and that, in turn, the autonomic management system could identify actions to take per component. However, these independent sets of policies may possibly yield multiple directives from which the autonomic manager must select one or more appropriate actions. In this work, we look at an approach which tries to select one or more actions based on previous behavior. Previous decisions are captured in the form of states and transitions which in turn are used to try to find good sequences of actions. We describe how the states are formed and actions identified. We describe the implementation of an autonomic manager using this approach and report on the experience in managing a dynamic Web server.