Simulating Taverna workflows using stochastic process algebras

Scientific workflows provide powerful middleware for scientific computing in that they represent a central abstraction in the research task, by simultaneously acting as an editable action plan, collaboration tool, and executable entity. Taverna workflows in particular have been widely accepted in the bioinformatics community, due to their flexible integration with web service analytical tools that are the essential tools of any bioinformatician. However, the semantics of Taverna has so far only been qualified in terms of the functional composition and data processing. While correct, and useful for reasoning about functional and trace equivalences, this aspect does not help with modelling the throughput and utilisation of individual services in the workflow. In this paper we present a stochastic process model for Taverna, and use it to perform execution simulations in Microsoft’s SPIM tool. The model also opens up the possibilities for further static analyses that are explored.