Sara Migliorini 1 , †

Introduction Scientific Workflow Management Systems (WfMSs) are software systems developed for automating scientific experiments that need to deal with huge amounts of data. The main goal of these systems is to facilitate the reuse and integration of domain specific functions and tools through a graphical environment. Scientific WfMSs can be used to automate repetitive error-prone activities, such as data access, integration, transformation, analysis and visualization, allowing scientists to focus on the domain specific aspects of their work. At the same time, they optimize workflow execution and resource consumption in a transparent way, scheduling the work on distributed architectures, such as a cluster of computers or a Grid environment. The effectiveness of a scientific WfMS mainly depends on the available functions for a particular scientific domain and the integration with domain-specific tools. The diverse nature of tasks being performed across different scientific domains (e.g. biology, physics) has led each scientific community to develop and adopt a new workflow solution best suited for their requirements, rather than extending an existing solution [1]. As a result, to date there is a variety of offerings and still no accepted standard scientific WfMS. In light of this, in this paper we aim to provide a comparative analysis of different scientific WfMSs. The purpose is to evaluate their suitability independently from the available domain-specific functions. In the Business Process Management (BPM) community, the Workflow Patterns Initiative [2] provides a framework to compare the suitability of WfMSs for business process design based on a set of recurring features (so-called patterns), as they are provided by various WfMSs. Inspired by this initiative, we present a pattern-based evaluation of three well-known open-source scientific WfMSs: Kepler [3], Taverna [4] and Triana [5]. In doing so, we also compare them with traditional WfMSs. In particular, we discuss why some groups of patterns are directly supported by business WfMSs but not by scientific WfMSs. For example , we observed that in some cases a pattern is not directly supported because it is deemed not to be relevant for that application domain; in other cases this is a reflection of the immaturity of the system. In order to conduct