Quality of Service Provision for Grid Applications via Intelligent Scheduling

Grid computing can be defined as coordinated resource sharing and problem solving in dynamic, multi-institutional collaborations. The success of a Grid infrastructure is based on a number of fundamental requirements, including the ability to provide dynamic and efficient services. Underpinning such a system is the need to ensure that the Grid infrastructure is delivering the required Quality of Service to its users. Quality of Service (QoS) is the ability of an application to have some level of assurance that users' requirements can be satisfied. It can be seen as an agreement between a user and a resource provider to execute an application within a guaranteed time frame at a pre-specified cost. As a rule, the higher the cost paid by the user, the smaller the execution time a resource provider can ensure. The novel contribution of the proposed project is to produce a new type of Grid resource broker with an advanced scheduling component aimed at optimising both resource usage costs and applications' execution times to enforce QoS. It should combine the features of two types of brokers: system-centric and user-centric providing a transparent means of meeting users' requirements and at the same time optimising the usage of Grid resources on the provider's end. This proposal is timely in that it addresses the need for continued development of infrastructure support for Grid computing. It responds to the increased attention of the Grid community to QoS provision and higher expectations of Grid users to receive adequate services at an agreed price payable for the agreed execution time. Our research will take advantage of the achievements in the classical scheduling theory and the newly emerged Grid scheduling research and will advance the frontiers of both areas. Grid applications give rise to new enhanced scheduling models. These enhanced models generally cannot be handled by the existing scheduling techniques developed mainly for manufacturing applications. They are characterised by complex additional constraints including those related to data storage and data transfer, co-ordinating the execution of linked tasks and arranging the required data interchange. Further challenges are related to the dynamic nature of Grid systems with the changing availability and quality of resources. The new aspects of QoS provision introduce additional complexity to scheduling. The project will draw on expertise of two established research groups at the University of Leeds: Algorithms and Complexity Group and Collaborative Architectures and Performance Group. The Algorithms and Complexity Group performs multidisciplinary research in algorithms, combinatorics and optimisation. Inter alia, the group develops and analyses advanced mathematical techniques for solving complex optimisation problems including those related to the areas of scheduling and optimal resource allocation. Research of the Collaborative Architectures and Performance Group focuses on Intelligent Infrastructures for large-scale applications. In particular, research of the group brings together e-science, Grid and adaptive computing systems research.