Scalable, fault-tolerant management of grid services: Application to messaging middleware

The service-oriented architecture has come a long way in solving the problem of reusability of existing software resources. As Service-based architectures emerge, management of the application which comprises of a large number of distributed services becomes even more difficult as resources appear, move and disappear across the network. Further, the application components may span disparate network boundaries, which add a variety of constraints such as network policies, blocked transports and authentication requirements. Services exist on different platforms and are written in different languages. This makes use of any single management technology inefficient and promotes non-interoperability.; In this thesis, we present a management architecture that combines publish-subscribe and service-oriented computing principles for managing a set of distributed entities. The use of service-oriented architecture adds interoperability to the management process. The proposed system adopts a distributed hierarchical architecture to achieve scalability. We show that the architecture is tolerant to failures in the management framework itself and can be extended to provide userlevel fault-tolerance of managed resources by implementing appropriate policies. Finally, we present an empirical evaluation of the system and demonstrate that the proposed architecture adds an acceptable number of additional resources required for providing fault-tolerance of various components in the system.