부산대학교 도서관

Wide area networks (WANs) are increasingly relying on software-defined networking to orchestrate data transmission. Google and Microsoft built B4 and SWAN, respectively, to interconnect their data centers in WANs, achieving higher network utilization and lower delay, where different priorities can be applied for bandwidth allocation and route selection in their centralized traffic engineering. However, simply relying on the priorities cannot accurately capture the complex properties of the quality of experience (QoE) function corresponding to the end users and always result in a suboptimal solution. We propose QTE, a novel traffic engineering in software-defined WANs, which proactively enforces forwarding policies by coordinating the traffic demand of each data center. The goal of QTE aims to maximize end-user QoE for client-trigger traffic and maximize network throughput for background traffic. We make several technical contributions in designing QTE. First, we formulate maximizing end-user QoE problem as two optimization programs and propose a heuristic algorithm to solve the problem. We further present a concrete design and implementation of the system based on the Floodlight controller. Extensive experiments use Openflow switches on Mininet and numerical simulations, which shows that QTE increases end-user experiences by 45.3% compared with prior work and can be readily implemented on the Floodlight controller.