Global E-Science Collaboration

Abstract

Today's e-science, with its extreme-scale scientific applications, marks a turning point for high-end requirements on the compute infrastructure and, in particular, on optical networking resources. Although ongoing research efforts are aimed at exploiting the vast bandwidth of fiber-optic networks to both interconnect resources and enable high-performance applications, challenges continue to arise in the area of the optical control plane. The ultimate goal in this area is to extend the concept of application-driven networking into the optical space, providing unique features that couldn't be achieved otherwise. Many researchers in the e-science community are adopting grid computing to meet their ever-increasing computational and bandwidth needs as well as help them with their globally distributed collaborative efforts. This recent awareness of the network as a prime resource has led to a sharper focus on interactions with the optical control plane, grid middleware, and other applications. This article attempts to explain the rationale for why high-end e-science applications consider optical network resources to be as essential and dynamic as CPU and storage resources in a grid infrastructure and why rethinking the role of the optical control plane is essential for next-generation optical networks.