Bandwidth scheduling with multiple fixed node-disjoint paths in high-performance networks

Abstract

Many large-scale applications generate large volumes of data that must be transferred over high-performance networks for various storage or analysis purposes. Such requirements call for a fast bandwidth scheduling solution to discover feasible and efficient reservation options in a time-varying network environment. We investigate a bandwidth scheduling problem with two node-disjoint paths, referred to as BS-2NDP, to support big data transfer. In BS-2NDP, we further consider two different types of paths: (i) two fixed paths of fixed bandwidth (2FPFB), and (ii) two fixed paths of variable bandwidth (2FPVB). We show that both 2FPFB and 2FPVB are NP-complete, and then design heuristic approach-based solutions, which are implemented and tested in both simulated and real-life networks. Extensive results illustrate that the proposed heuristics achieve a close-to-optimal performance in small-scale networks, and significantly outperform other heuristic approaches in large-scale networks.