An Ontology Model for Execution Records of Grid Scientific Applications

Abstract

Records of past application executions are particularly important in the case of loosely-coupled, workflowdriven scientific applications which are used to conduct in silico experiments, often on top of Grid infrastructures. In this paper, we propose an ontology-based model for storing and querying over the execution records of Grid scientific applications. The model comprises various aspects of the execution, including experiment, performance, resource and provenance information. In addition, domain ontologies of data and applications are associated with the generic execution entries to capture the semantics of data and computations used in the experiment. We introduce a hierarchy of monitoring events for application execution and describe the process whereby high-level records are built from low-level monitoring events. We demonstrate how the proposed model is useful for end-user oriented, ontology-based querying over application execution records, and how ontologies enable semantic information integration between experiment data and execution metadata, enabling queries otherwise not possible.