This thesis presents a domain-independent approach to dynamic system management using goals and policies. A goal is a general, high-level aim a system must continually work toward achieving. A policy is a statement of how a system should behave for a given set of detectable events and conditions. Combined, goals may be realised through the selection and execution of policies that contribute to their aims. In this manner, a system may be managed using a goal-directed, policy-based approach. The approach is a collection of related techniques and tools: a policy language and policy system, goal definition and refinement via policy selection, and conflict filtering among policies. Central to these themes, ontologies are used to model application domains, and incorporate domain knowledge within the system. The ACCENT policy system (Advanced Component Control Enhancing Network Technologies, http://www.cs.stir.ac.uk/accent) is used as a base for the approach, while goals and policies are defined using an extension of APPEL (Adaptable and Programmable Policy Environment and Language, http://www.cs.stir.ac.uk/appel). The approach differs from existing work in that it reduces system state, goals and policies to a numerical rather than logical form. This is more user-friendly as the goal domain may be expressed without any knowledge of formal methods. All developed techniques and tools are entirely domain-independent, allowing for reuse with other event-driven systems. The ability to express a system aim as a goal provides more powerful and proactive high-level management than was previously possible using policies alone. The approach is demonstrated and evaluated within this thesis for the domains of Internet telephony and sensor network/wind turbine management.
Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:513785 |
Date | January 2008 |
Creators | Campbell, Gavin A. |
Contributors | Turner, Kenneth J. : Magill, Evan H. |
Publisher | University of Stirling |
Source Sets | Ethos UK |
Detected Language | English |
Type | Electronic Thesis or Dissertation |
Source | http://hdl.handle.net/1893/1220 |
Page generated in 0.0014 seconds