Service-based architectures represent the next evolutionary step in the development of e-science, namely, the transformation of the Internet from a commercial marketplace to a mechanism for sharing multidisciplinary scientific resources. Although scientists in many disciplines have become increasingly reliant on distributed computing technologies for data processing and dissemination, the record of the processing history and origin of a data product, that is its data provenance, is often nonexistent, incomplete or impossible to recover by potential users. This thesis aims to address data provenance issues in service-based environments, particularly to answer how a scientist who performs a workflow execution in such an environment can (1) document the data provenance for a data item created by the execution, and (2) use the provenance documentation as a recipe to re-execute the workflow. This thesis pro poses a provenance model for delivering data provenance support in a service-based environment. Through the use of an example scenario of a scientific workflow in the Astrophysics domain, we explore and identify components of the provenance model. The provenance model proposes a technique to collect and record data provenance for service-based workflow executions. The technique facilitates the collection of data provenance of workflow execution at runtime. In order to record the collected data provenance, the thesis also proposes a specification to represent provenance to de scribe the processing history whereby a piece of data was derived. The thesis also proposes query interfaces that allow recorded provenance to be queried, has formulated a technique to construct provenance graphs, and supports the re-execution of past workflows. The provenance representation specification, the collection technique, and the query interfaces have been used to implement a prototype system to demonstrate the proposed model. The thesis also experimentally evaluates the scalability of the components implemented.
展开▼
