Open Access Dissertation
Date of Award
In recent years, significant scientific advances are increasingly achieved through complex scientific processes. As the exponential growth in computing technologies and scientific data, a scientific workflow may comprise a large number of heterogeneous scientific services and applications, provided by different organizations. These services, applications, and their associated data are usually distributed across heterogeneous computing environments. The integration and management of such scientific workflows are pushing the limits of current workflow technology. This dissertation presents an integrated solution to composing, scheduling, executing and developing scientific workflows and scientific workflow management systems.
To provide a foundation for workflow composition, scheduling, execution and management, we propose the first reference architecture for scientific workflow management systems. The reference architecture not only provides a high-level organization of subsystems and their interactions in a workflow system, but also provides a basis for comparison between different systems and a guidance for the architectural design of an SWFMS in a specific scientific domain. To integrate heterogeneous services and applications and enable them composed to workflows, we propose a task template model which not only provides an appropriate abstraction of heterogeneous services and applications, but also encapsulates the composition and mapping of shims and functional task components within a task interface. Our proposed task specification language (TSL) not only integrates heterogeneous services and applications into uniform workflow tasks, but also provides a solution to address both TYPE-I and TYPE-II shimming problems in composing scientific workflows. To schedule scientific workflows in emerging services computing environments, we propose two workflow scheduling algorithms, the SHEFT algorithm and the SCPOR algorithm, to prioritize tasks in a workflow, map tasks onto suitable resources and order the execution of tasks on the assigned resources, so that the workflow makespan can be minimized. Our extensive experiments have shown that our proposed algorithms not only outperform other algorithms for large-scale, data-intensive and compute intensive workflows, but also allow the assigned resources elastically change on demand of the size of workflows. To execute workflows on distributed computing environments, we propose a task run model to model the run-time behaviors of tasks. The proposed task run description language (TRDL) enables the execution of task instances constructed from heterogeneous services and applications. We also develop an SOA based task management subsystem to manage all task templates, task instances and task runs for the invocation and execution of various heterogeneous task components. Finally, our developed SOA based workflow management system, the VIEW system, and a VIEW based workflow application system, the FiberFlow system, validate our architectures, models, languages, and algorithms.
Lin, Cui, "Scientific Workflow Integration For Services Computing" (2010). Wayne State University Dissertations. Paper 19.