Abstract

Very large scale computations are now becoming routinely used as a metliodology to undertake scientific research. In this context,)provenance systems ’ are regarded as the equivazent of the scientist’s logbook far in silico experimenlation: provenance captures the documentation of the process that led to some result. Using a protein compressibility analysis application, we derive a set of generic use cascs for Q provenance system. In order to support these, we address the following fundamental questions: what is provenance? how to record il? what is the performance impacr for grid execution? what is the performance of reasoning? In doing so, we dejne a technologyindependent notion of provenance that captures intemctions between components, internal componenr information and grouping of interactions, so as to allow us to analyse and reason about rhe execution of scientific processes. In order to siipport persistennl provenance in heterogeneous upplications, we introduce a separate provenance store, in which provenance documentation can be stored, archived and queried indepsndently of the technology used to mn the application. Through a series of practical tests, we evaluale the pe~onnance impact of such a provenance system. In summary, we demonstrate that provenance recording overhead of our protolype system remains under 10 % of execution rime, and we show that the recorded information successfully supports our use cases in U perjbmant mannel: 1

Citations