Automating Performance Analysis from Taverna Workflows

Workflow systems provide support for combining components to achieve a particular outcome. Various approaches from software engineering have been utilized within such systems, such as the use of design patterns to support composition, and the use of a software engineering lifecycle to support workflow construction and execution. As components used within a workflow may be implemented by third parties, it is often necessary to be able to determine the impact a particular component composition will have on the overall execution of a workflow. A method for predicting the execution time of a given workflow is proposed. First, the method obtains a model from a given workflow in an automated way. The model obtained is a Reference net – a specific type of Petri net. Features of Reference nets can subsequently be exploited, such as the possibility of building hierarchical workflow models which can facilitate the modelling process. The Reference nets are extended so that each task in the model is parameterised with a time value, representing the execution time of the task. We propose several timing profiles: those obtained from real measurement of the workflow system, from stochastic and constant values which allow us to test the model behaviour under specific situations.