A typical large-scale, data driven, processing pipeline consists of a great amount of generic components that may be reusable inside other applications. However, there is currently no programming model that makes it possible to reuse existing data driven software components in a cluster-based setting. Offering such a model would promote the reuse of existing software, saving development time. Our work introduces the notion of reactive workflows, a technique that combines concepts from scientific workflows and reactive programming. We investigate the required abstractions to create such a workflow and offer these in a domain specific language, called Skitter. We plan to evaluate this framework based on its expressivity and horizontal scalability.