Towards Use-Case Driven Self-Management of Distributed Systems

Abstract

The growing complexity of distributed systems makes their management a challenge. Manual management of these systems is costly and time consuming. Autonomic computing has been proposed to reduce the cost of maintaining complex systems by developing computer systems capable of self-management. A self-managed system consists of multiple autonomic managers and multiple management units. This thesis explores the potentials of autonomic computing for management of existing distributed systems. The management units in the autonomic approach proposed in this thesis are system behaviours. For a self-managed system to be able to recognise and solve a large portion of its malfunctions on its own, it needs to have knowledge about its own behaviours. As this knowledge is mainly specified in use-case notations during the software design and development phase, this thesis proposes to reuse the use-case notations for self-management purposes. Use-cases are usually expressed in a semi-formal way. They must be expressed in a formal language to be processable and understandable by software modules (autonomic managers). For this purpose, a management model (the way an autonomic manager manages a use-case realisation and the way multiple autonomic managers cooperate with each other) and a system model (the way a managed system provides its functionalities and description of the internal structure of a managed system) are proposed and constructed. All knowledge included in these generic models is represented in formal knowledge representation languages (the Semantic Web languages OWL and SWRL). A self-management framework based on the management and system models is designed and implemented. The framework provides an execution environment for autonomic managers to manage behaviours of an existing distributed system. Furthermore, the framework can be used to generate code for the autonomic managers, based on declarative behaviour descriptions.