Provenance based diagnosis for scientific workflows

Scientific research has become collaborative activities among scientists. Many scientists share data and scientific workflows with each other to facilitate scientific research. However, during using shared scientific workflows, it's hard for scientists to figure out suspicious data and processes, i.e., scientific workflow diagnosis. In this paper, we propose an approach for diagnosing scientific workflows based on the data provenance technology and reusing of diagnosis information. The main innovation of this paper is twofold: 1) By using data provenance technology, we deduce integrated diagnosis information of scientific workflows with the help of scientists' understanding of data; 2) We propose a novel model to preserve diagnosis information, and design a method to reuse the diagnosis information. We implement a prototype system based on the Kepler scientific workflow management system.