Provenance-based Validation of E-Science Experiments

Download

PDF Download (179Kb)

Description/Abstract

E-Science experiments typically involve many distributed services maintained by different organisations. After an experiment has been executed, it is useful for a scientist to verify that the execution was performed correctly or is compatible with some existing experimental criteria or standards. Scientists may also want to review and verify experiments performed by their colleagues. There are no exsiting frameworks for validating such experiments in today's e-Science systems. Users therefore have to rely on error checking performed by the services, or adopt other ad hoc methods. This paper introduces a platform-independent framework for validating workflow executions. The validation relies on reasoning over the documented \emph{provenance} of experiment results and \emph{semantic descriptions} of services advertised in a registry. This validation process ensures experiments are performed correctly, and thus results generated are meaningful. The framework is tested in a bioinformatics application that performs protein compressibility analysis.