A scalable solution to provide Quality of Service in Next-Generation Networks

Next-Generation Networks (NGNs) will support Quality of Service (QoS) for multimedia traffic over an IP-based infrastructure. A scalable solution to provide also stringent guarantees is required. DiffServ architecture can offer different levels of service at low complexity, but it is not basically able to efficiently provide end-to-end absolute QoS for real-time traffic. Many research activities have addressed this issue, proposing either an absolute or a relative approach. While the former is complicated to implement in the global Internet, the latter is simpler and can be easily realized without arising scalability concerns because exploiting the Proportional Differentiation Model (PDM), in which the performance distance between classes is proportional to quality differentiation parameters that Network Service Provider (NSP) can configure. This work aims to achieve absolute delay guarantees relying on a PDM that can be easily deployed in DiffServ architecture using a proportional scheduler, like Advanced Waiting Time Priority (AWTP). The key idea is to enhance the end-to-end delay differentiation provided by PDM, with a run-time class adaptation, which dynamically assigns the service class to critical traffic in order to fulfil its end-to-end delay requirements. Simulation results have been collected to analyse the trade-off between the fast reaction to load changes and the system stability with different measurement processes, employed to decide for a class promotion or downgrade. The validity and good performance of our proposal have been demonstrated over various scenarios also in the critical case of network congestion.