NeTS-NR: Traffic Quantization: A Formal Framework for QoS and Scalability in Packet-Switched Networks

North Carolina State University, Raleigh NC

Investigators

Abstract

The goal of this project is to develop traffic quantization techniques for addressing the scalability challenges in packet-switched networks, as well as prototype implementations of protocols and algorithms for quantized networks. Traffic quantization is a new method for supporting per-flow functionality, which can alleviate the complexity associated with per-flow QoS. A quantized network offers only a small number of service levels. Rather than receiving the service requested, each flow is mapped to one of the levels offered by the network; this mapping is performed in a way that guarantees a QoS at least as good as that requested by the flow. The directional p-median problem has been introduced as a formal framework for reasoning about the bandwidth quantization problem, but there is a new set of research avenues to pursue. Specific objectives include the study of stochastic quantization and quantization for a vector of traffic parameters, and the implementation of efficient algorithms (scheduling, bandwidth-guaranteed routing) for quantized networks. Furthermore, this project will investigate quantization as a technology enabling scalability in a range of applications and problem domains (multiprocessor scheduling and Grid computing). The end-result will be a collection of traffic quantization techniques with formally verified properties, and a suite of scalable control and data plane protocols and algorithms to support QoS guarantees in quantized networks. It is expected that the results of this work will address the QoS and scalability problems facing the Internet, and will lead to the development of novel network architectures with the ability to cater to very large sets of heterogeneous users by offering and guaranteeing a menu of optimized services.