Development of Dynamic Meter-Based Rate Limiter in SDN for Optimizing Multimedia Traffic Using P4 Language

Abstract

Software Defined Networking (SDN) is a new paradigm in the networking environment that decouples the control plane from the data plane and introduces network agility and flexibility. This new technology has paved the way for innovative solutions to tackle the complexities of providing efficient Quality of Service (QoS) for multimedia traffic, which has grown exponentially at the same pace as the internet and internet-connected devices. Existing methods rely on legacy devices where QoS performance is poor due to their static metering and packet sampling approach. Previous researchers made significant contribution by introducing the Meter Band Rate Evaluation (MBE) mechanism which incorporates Band Rate Description Language (BRDL) and Band Rate Evaluator (BRE) components directly in the data plane over OpenFlow-based SDN. However, while MBE method effectively improved QoS for mission-critical traffic, it remained limited by the OpenFlow’s configuration capabilities. This work leverages the new programming protocol-independent packet processors (P4) language to develop an indirect meter linked to the SDN Match-Action table with an efficient algorithm for dynamic, real-time adjustments of threshold rates and packet sampling interval. The proposed algorithm also integrates a packet recirculation strategy to manage oscillations in high-priority traffic, while low-priority traffic is distributed across N-distinct routes to maintain overall network efficiency. A genuine datacenter fabric topology is used to test the performance. The outcome of the experiments demonstrates the success of the approach in three out of the four metrics compared with MBE by maximizing throughput up to 30.14%, minimizing jitter by 46.93%, and reducing packet loss rate by 61.42%.