Automated provenance collection for cca component assemblies

Abstract

The problem of capturing provenance for computational tasks has recently received significant attention, due to the new set of beneficial uses (for optimization, debugging, etc.) of the recorded data. We develop a provenance collection system aimed at scientific applications that are based on the Common Component Architecture (CCA) that alleviates scientists from the responsibility to manually instrument code in order to collect provenance data. Our system collects provenance data at the granularity of component instances, by automatically recording all method invocations between them, including all input and output parameters. By relying on asynchronous communication and using optimizations to handle large data arrays, the overhead of our system is low-enough to allow continuous provenance collection. © 2009 Springer Berlin Heidelberg.