ABSTRACT
This chapter presents the computing node network energy consumption problem of the most famous fat-tree interconnection networks for low communication to computation ratio parallel application. It presents a heuristic network energy-efficient workflow task scheduling algorithm, which includes task level computing, task subdeadline initialization, task dynamic adjustment, and a data communication optimization scheme among tasks. In general, high-performance scientific parallel applications usually consist of a set of precedence constrained tasks. In fact, data communication inevitably exists in scientific parallel applications. The pseudo-code of the heuristic network energy-efficient workflow scheduling algorithm (NEEWS). The chapter presents the computing nodes and high-speed interconnection networks of the systems, parallel application workflow DAG model, and basic concepts about energy consumption.
