TrafficVision: A Case for Pushing Software Defined Networks to Wireless Edges

摘要: In wireless network edges, knowing the flow types and applications can enable various policy-driven managements (i.e. traffic offloading, BYOD, E2E QoS etc.). However, applying policies at links between mobile devices access points (APs) requires greater visibility control on generated from devices. The recent advent of Software-Defined Networking (SDN) could fine-grained management edge. in existing solutions, SDN uses external Deep-Packet Inspection (DPI) engine that additional potentially heavy loaded computational resources to perform packet analysis. Moreover, become more dynamic (rapid install/update), diverse complex (individual generate multiple types) which scalability granularity requirements challenging current DPI solutions. addition, is unreliable classifying application's encrypted packets. Therefore, this paper we present design development TrafficVision extends SDN's layer architecture have real-time policy making edges. More specifically, carefully framework develop tools allow scalable, efficient flexible way classify flows fashion using Machine-Learning (ML) based technique. We evaluate our system performance CPU utilization, overhead throughput metrics. Finally, as a proof concept, simple case study application exploits TrafficVision.