MAGNET : a dynamic resource management architecture

This thesis proposes a new dynamic resource management architecture, Magnet , to meet the requirements of users in exible and adaptive systems. Computer systems no longer operate in centralized isolated static environments. Technological advances, such as smaller and faster hardware, and higher reliability of networks have resulted in the growth of mobility of computing and the need for run-time reconngurability. The dynamic management of this diversity of resources is the central issue addressed in this thesis. Applications in environments with frequently changing characteristics are required to participate in dynamic resource management , to adapt to ever-changing conditions, and to express their requirements in terms of quality of service. Magnet enables dynamic trading of resources which can be requested indirectly by the type of service they ooer, rather than directly by their name. A dedicated component, the Trader, matches requests for services against demands and establishes a component binding resource allocation. In addition, the architecture is extensible it does not constrain the information on services and allows user-customization of the matching process. Consequently, this allows resource dee-nitions to be parametrized (to include QoS-based characteristics), and the matching process to be user-customized (to preform QoS-based negotiation). In order to ful-ll the requirements of users relying on ever-changing conditions, Magnet enables runtime adaptation (dynamic rebinding) to changes in the environment, constant monitoring of resources, and scalability of the architecture. The generality of the Magnet architecture is illustrated with several examples of resource allocation in dynamic environments. iii Acknowledgements Over the last three years, there were many people who have directly, or indirectly, intentionally or accidently contributed to this work having come to fruition. They are all owed my thanks. First of all, I would like to thank all my oocial and unoocial supervisors: thèfather' of the component-based resource management idea, gave me expert technical support in various aspects of this thesis, and desperately needed encouragement before public presentations at my rst conferences. Peter Osmon gave me constant encouragement and support in his ever-optimistic style. His guidance and helpful comments, particularly during the writing-up phase of my thesis, are much appreciated. Many thanks are due to Steve Crane, who appeared in my nal year, just in time to suggest Regis for my implementation environment. His perfectionist attitude, insight into the subject, and expert technical support helped me to improve, justify, and better formalize this work. My thanks are also due to Julie McCann for …