System management research via behavior characterization

Since system management is very much environment and application dependent, past system management research has dealt mainly with solving problems of specific systems. As a result, the research results obtained often can not be reused or extended to different environment and applications. To remedy such shortcomings, we propose to study the behavior of system management techniques in generic environments and applications. Based on the abstract representations of the process, operating environment and application, a rapid prototype or testbed can then be constructed to represent the system operating environment and its application. System management techniques are categorized into classes according to their similarity. The behavior of these classes of techniques can then be determined on the prototype or testbed. Examples are given to illustrate the process of the proposed methodology.<<ETX>>