IWIR: a language enabling portability across grid workflow systems

Abstract

Today there are many different scientific Grid workflow management systems using a wide array of custom workflow languages. Some of them are geared towards a data-based view, some are geared towards a control-flow based view and others try to be as generic, and therefore often complex, as possible. All of these languages and custom workflow management system front-ends fulfill special needs and workflow creation paradigms for their respective user communities. The problem is that once a workflow application has been created in one of these systems, it becomes very hard to share the workflow with users working with different systems. Portability and interoperability between current systems barely exists. In this work, we present a common workflow language for use as an intermediate exchange representation by multiple workflow systems. It comprises atomic tasks, compound tasks including conditionals, sequential and parallel loops as well as an expressive set of data types and data flow constructs.