Traffic Quantization in Packet-Switched Networks

We present traffic quantization as a new method for supporting per-flow functionality in packet-switched networks in an efficient and scalable manner. A quantized network offers a small set of service levels (tiers), and each flow is mapped to one of the service levels so as to guarantee a QoS at least as good as that requested by the flow. Traffic quantization trades off a small amount of system resources for simplicity in core network functions, including packet scheduling, traffic policing, etc. Our focus in this paper is on determining the service levels so as to minimize the amount of additional resources required by a quantized network. We develop a framework for addressing the traffic quantization problem by formulating it as a variant of the -median problem under a new distance measure, the directional rectilinear distance. We present an efficient optimal algorithm for the directional -median problem on the real line, and a robust heuristic for the same problem in multiple dimensions; these two problems correspond to quantization of a single and multiple traffic parameters, respectively. Our main finding is that, by appropriately selecting the set of service levels, the benefits of quantization can be achieved with only a small sacrifice in network resources.