User-centric traffic optimization in residential software defined networks

Abstract

Software defined networking (SDN) provides a centralized control framework with real-time control of network components, residential customer routers in particular, that allows automated per-user bandwidth allocation. However, employing dynamic traffic shaping for efficient bandwidth utilization among residential users is a challenging task. In this context, understanding application usage requirements for each individual user and translating them into network policies requires expertise beyond most residential users. This paper proposes a user-centric traffic optimization scheme by profiling users based on their application trends recorded using generic NetFlow records, in order to provide a better view of per user utilization. We also propose an SDN traffic monitoring and management application for implementing Linux-based hierarchical token bucket (HTB) queues customized for individual user profiles in real-time, according to user-defined priorities. The traffic management scheme scales well under both upstream and downstream network congestion by dynamically allocating dedicated bandwidth to users based on their profile priority, resulting in a decreased packet loss and latency for a selected set of high priority users.