Quality of service management in service-oriented grids

Grid computing provides a robust paradigm for aggregating disparate resources in a secure and controlled environment. The emerging grid infrastructure gives rise to a class of scientific applications and services in support of collaborative and distributed resource-sharing requirements, as part of teleimmersion, visualization and simulation services. Because such applications operate in a collaborative mode, data must be stored, processed and delivered in a timely manner. Such classes of applications have collaborative and distributed resource-sharing requirements, and have stringent real-time constraints and quality-of-service (QoS) requirements. A QoS management approach is therefore essential to orchestrate and guarantee the interaction among such applications in a distributed computing environment. Grid architectures require an underpinning of QoS support to manage complex computation-intensive and data-intensive applications, as current grid middleware solutions lack QoS provision. QoS guarantees in the grid context have, however, not been given the importance they merit. To enhance its functionality, a computational grid must be overlaid with an advanced QoS architecture to best execute those applications with real-time constraints. This thesis reports on the design and implementation of a software framework, called Grid QoS Management (G-QoSm). G-QoSm incorporates a new QoS management model and provides a service-oriented QoS management approach that supports the Open Grid Service Architecture. Its novel features include grid-service discovery based on QoS attributes, immediate and advance resource reservation, service execution with QoS constraints, and techniques for QoS adaptation to compensate for resource degradation, and to optimise resource allocation while maintaining a service level agreement. The benefits of G-QoSm are demonstrated by prototype test-beds that integrate scientific grid applications and simulate grid data-transfer applications. Results show that the grid application and the data-transfer simulation have better performance when used with the proposed QoS approach. QoS abstractions are presented for building QoS-aware applications, in the context of service-oriented grids. These abstractions are application programming interfaces to facilitate application developers utilising the proposed QoS management solution.