RABR: a service based on adaptive rate guarantees for real-time video in ABR networks

A new renegotiation service renegotiated ABR (RABR) and a dynamic bandwidth allocation strategy are presented for real-time video transfer in ABR networks. Static allocation services fail to meet both QoS and utilisation requirements. The proposed framework uses an adaptive control-theoretic approach for source buffer occupancy tracking to compute renegotiated rates. Trace-driven evaluation results show that a self-tuning controller is able to track encoder bit-rate changes and the closed-loop system is stable. We also present network simulation results that demonstrate excellent performance in reaching probabilistic QoS and utilisation targets. We believe the approach has more built-in flexibility than other renegotiation services, allowing the source rate to be modulated not only according to its needs but also in reply to network feedback. Overall, we show how to provide high-end video support on top of a static ABR service