See our new website at research.ict.csiro.au

Self-tuning End-to-End Quality of Service (QoS) for IP Networks

A new generation of high performance applications are being introduced to today’s IP networks that have fundamental requirements for the effective provisioning of quality of service (QoS). Such applications include Voice over IP (VoIP), interactive video/audio streaming, IP TV, distributed collaborative and learning environments to name a few. As these applications penetrate the global Internet, an effective means of providing end-to-end QoS will become more important for service providers who wish to access the revenue that these applications can deliver.

The current IETF Standards for delivering QoS, namely DiffServ and IntServ, are respectively limited by functionality and scalability and hence have not been very successful in wide scale global deployment. An ideal QoS-enabled Internet should provide the following features:

• End-to-end QoS guarantee
• Scalability with respect to network size and number of traffic flows
• Simple configuration at the network edge where end-to-end QoS guarantee are specified
• Adaptive bandwidth reservation and residual bandwidth allocation based on time-varying application demands and network routing
• Applicability to flows across multiple ISP domains
• Fair bandwidth allocation amongst all flows

We are currently investigating and implementing on network processor units (NPU) a scalable, distributed and adaptive protocol called RMP (rate management protocol) that addresses the current solutions’ shortcomings and will ultimately enable the ideal QoS support required for future high-performance IP networks.