Designing Workflow Components for e-Science

In this paper we present a general domain for the analysis of workflows and workflow components based on the notion of a collection of Turing machines sharing a set of tapes. We show that computationally equivalent workflows can be evaluated in terms of two dimensions: data complexity and process complexity. We show that this approach allows for the evaluation of various workflow architectures. Using this formal framework we prove that maximal simplicity, generality and consistency are mutually exclusive. Simplicity of and generality of workflow components leads to complexity of data structures and computational processes. This is an issue that deserves more attention from designers and users of workflow communication protocols. We define a formal version of the General Workflow Design Problem and show that this problem is decidable in the case of a finite number of topologies. Thus, automatic composition of workflows is possible in limited domains. Decidability for an infinite number of topologies remains an open question. We show how our findings from the formal framework manifest themselves in real world e-Science workflow environments.