Towards Quality of Service in Scientific Workflows by Using Advance Resource Reservations

The increasing interest in Web services has led to a recent proliferation of service-oriented platforms for supporting Scientific Workflows in a variety of e-Science domains. These platforms mainly address the application-specific composition and enactment of workflows. However, since scientific workflows are characterized by complex computations and large volumes of data, non-functional criteria like Quality of Service (QoS) guarantees can significantly raise the predictability of the computations, and thus constitute an important benefit to users. This paper introduces the DWARFS vision of a distributed execution engine for Scientific Workflows which features support for QoS guarantees by making use of Advance Resource Reservations. In particular, we present the DWARFS methodology to enact workflows under QoS constraints and report on the ongoing implementation activities.