Prescriptive provenance for streaming analysis of workflows at scale

摘要: We extend our approach capturing and relating the provenance performance metrics of computational workflows as a diagnostic tool for runtime optimization placement. One important challenge is volume extracted data, both provenance, even when specifying filters focusing on quantities interest in simulation. reduce this data by performing anomaly detection streaming store detected anomalies, an we call prescriptive provenance. This paper discusses Chimbuko architecture enabling approach. present use protein structure propagation workflow based NWChemEx. are testing algorithms preliminary results here obtained with Local Outlier Factor. While scaling remains challenge, these show that robust analysis promising