Co-Scheduling Scientific Workflows in Elastic Optical Networks

Abstract

E-science supports inter-disciplinary research that requires processing highly data-intensive workflows. These workflows require compute resources for processing, and storage resources to save data generated on computation. Additionally, there may be a need to make this data available to researchers at distinct and disparate locations, which requires network resources. Current-generation networks cannot scale to meet the demands of tomorrow's unpredictable application workflow scenarios. Elastic Optical Network (EON) is a cutting-edge technology which supports more data and faster data, without the need to construct larger networks. The challenge of joint scheduling of computational, storage and EON network resources to e-science workflows is a complex problem known as co-scheduling. In this paper, heuristics are proposed to investigate sequential and parallel co-scheduling. Simulation results are presented to demonstrate the effectiveness of the proposed approaches.