Traffic Quantization and its Application to QoS Routing

Abstract

In today's high speed networks, the data transmission rate available is usually a multiple of some base rate. An example is that of the T-carrier circuits having transmission rates that are a multiple of the fractional T1 system which is 64 Kbps. The base rate usually depends on the characteristics of the data being transmitted such as audio, video or other types of data. Our objective is to design efficient heuristics to quantize the user traffic demands based on bandwidth into multiples of such base rates such that the excess bandwidth spent due to quantization is minimized. In doing so, our contribution is twofold. First, there is always a penalty that one has to pay to reconfigure the system for each transmission. This reconfiguration time could be due to several factors such as memory operations and table lookups. Our Heuristics try to minimize such control/management costs. Secondly, our schemes have the flexibility on the values that the base rate can take which is consequently not hardwired into the system. Our work is greatly influenced by the extensive study in the field of location theory. We achieve our goal by modifying the p-median problem and solving it efficiently using dynamic programming and exploiting the properties of totally monotone matrices. In spite of having an additional constraint of base rate to quantize the traffic, we show that our heuristics have similar bandwidth losses when compared to previous study in the area of traffic quantization. In addition, we save more on control and management costs which are incurred during the reconfiguration of each flow. We also demonstrate through simulations that the blocking probability, defined as the ratio of number of flows blocked due to unavailability of the residual bandwidth in the network to the total number of flows, using our scheme is comparable to that of the continuous network.