This paper explores how process pipeline scheduling may become a viable strategy for executing workflows. It first details a workflow optimization and execution algorithm that reduces runtime space. The optimization strategy pipelines the communication between as many processes as possible, within the bounds of the storage space available, and depends on generic properties of datasets and processes. Then, the paper proves that the process pipeline scheduling problem is NP-Complete. Finally, it presents a greedy process pipeline scheduling algorithm which has a viable performance.
Melissa Lemos, Marco A. Casanova, Antonio L. Furta