What makes scientific workflows scientific?

A scientific workflow is the description of a process for accomplishing a scientific objective, usually expressed in terms of tasks and their dependencies [5]. While workflows have a long history in the database community as well as in business process modeling (where they are also known as business workflows), and despite some early works on scientific workflows [3,10], the area has only recently begun to fully flourish (e.g., see [1,2,9,7,4,11]). Similar to scientific data management which has different characteristics from traditional business data management [8], scientific workflows exhibit new challenges and opportunities that distinguish them from business workflows. We present an overview of these challenges and opportunities, covering a number of issues such as different models of computation, scalable data and process management, and data provenance and lineage handling in scientific workflows.