A Framework for Grid-Enabling Scientific Workflow Systems. Architecture and application case studies on interoperability and heterogeneity in support for Grid workflow automation.

Abstract

Since the early 2000s, Service Oriented Architectures (SOAs) have played a key role in the development of complex applications within a virtual organization (VO) context. Grids and workflows have emerged as vital technologies for addressing the (SOA) paradigm. Given the variety of Grid middleware, scientific workflow systems and Grid workflows available, bringing the two technologies together in a flexible, reusable and generalized way has been largely overlooked, particularly from a scientific end user perspective. The lack of domain focus in this area has led to a slow uptake of Grid technologies. This thesis aims to design a framework for Grid-enabling workflows, which identifies the essential technological components, how these components fit together in layered architecture and the interactions between them. To produce such a framework, this thesis first investigates the definition of a Grid-workflow architecture and mapping Grid functionality to workflow nodes, focusing on striking a balance between performance, usability and the Grid functionality supported. Next, it presents an examination of framework extensions for supporting various forms of Grid heterogeneity, essential for ii VO based collaboration. Given the complex nature of Grid technologies, the work presented here investigates abstracting Grid based workflows through high-level definitions and resolution using semantic technologies. Finally, this thesis presents a way to resolves abstract Grid workflows using semantic technologies and intelligent, autonomous agents. The frameworks presented in this thesis are tested and evaluated within the context of domain-based case studies defined in the SIMDAT, BRIDGE and ARGUGRID EU funded research projects.