Resource Based Pricing Framework for Integrated Services Networks Mostafa H. Dahshan University of Oklahoma/Electrical and Computer Engineering, Tulsa, OK, U.S.A Email: mdahshan@ou.edu Pramode K. Verma University of Oklahoma/Electrical and Computer Engineering, Tulsa, OK, U.S.A Email: pverma@ou.edu Abstract—This paper addresses the impact of Quality of Service on resource requirements in networks that implement exclusive bandwidth allocation, such as IntServ. The paper proposes a framework for pricing flows based on the impact of their reservations on the resources for which the network must be provisioned. The developed framework is analytical and is based on the economies associated with aggregating vs. segregating exclusive bandwidths that cater to customers demanding a specified Quality of Service. Index Terms—QoS Pricing, Integrated Services, Resource reservation, Economies of scale, Segregated flows, Aggregated flows. I. INTRODUCTION The Internet was initially designed as a best-effort network in which all user traffic is treated equally. However, the diversity of application demands and users’ willingness to pay have made it imperative to develop techniques for providing a certain level of assurance of resource availability based on application or user requirements. This has stimulated the development of Quality of Service (QoS) techniques such as the Integrated Services architecture (IntServ) [1], the Differentiated Services architecture (DiffServ) [2] and Multi-Protocol Label Switching (MPLS) [3]. While a significant research effort has been put in the design of QoS mechanisms, no such attention has been given toward understanding the cost of guaranteed resource availability and its effect on service pricing. This has resulted in pricing schemes that do not necessarily map the cost to the price of providing the service. Various pricing schemes have been proposed ranging from the simplest flat-rate pricing [4, 5] to more sophisticated techniques such as usage-based pricing [6], priority based pricing [7, 8] and congestion-based pricing [9, 10]. Rather than proposing a specific pricing scheme, the purpose of this paper is to understand the actual cost of offering guaranteed resources to specific flows under different scenarios of traffic characteristics and to provide a framework that can be deployed in designing the appropriate pricing scheme depending on the traffic pattern in the target network. As shown in [11, 12], the performance of shared communication channels is improved when traffic flows having disparate parameters (packet size and arrival rate) are segregated. Similarly, the performance for homogeneous flows is improved when these flows are aggregated. As a result, it is expected that aggregating heterogeneous flows or segregating homogeneous flows would impose some penalty. In the former case, the penalty occurs in the form of higher delay and jitter. In the latter case, the penalty occurs in the form of inefficient usage of bandwidth. In this paper, the IntServ QoS model will be used to study the effect of bandwidth reservation. IntServ uses the RSVP protocol to allocate bandwidth for each flow upon connection setup. The delay and jitter performance are compared before and after reservation to calculate the additional cost incurred on the network. In order to obtain tractable result, the M/M/1 queueing model is used for most of the analytical work. While the M/M/1 model facilitates simple calculations, it has been shown that self- similar stochastic models provide more accurate characterization of the Internet traffic [13-16]. Some insights into the M/G/1 model are therefore provided in the appendices. The M/M/1 analyses are still necessary because they provide closed form representations that give better understanding of the cost requirements. In addition, it has been shown that the M/M/1 model is still applicable in heavy loaded networks [17]. The reminder of this paper is organized as follows: Related studies on pricing models are reviewed in Section II. The IntServ model is briefly described in Section III with emphasis on the Controlled Load (CL) service class. The economies of scale of segregation versus aggregation of flows are reviewed in Section IV. Sections V and VI provide delay and jitter analysis, respectively, for segregated versus aggregated homogeneous flows sharing a common pool of bandwidth. The results obtained in Based on “Pricing for Quality of Service in High Speed Packet Switched Networks”, by Mostafa H. Dahshan, and Pramode K. Verma which appeared in the Proceedings of the IEEE Workshop on High Performance Switching and Routing 2006, Poznan, Poland, June, 2006. © 2006 IEEE. 36 JOURNAL OF NETWORKS, VOL. 2, NO. 3, JUNE 2007 © 2007 ACADEMY PUBLISHER