A CLASS-BASED ADMISSION CONTROL ALGORITHM FOR SHARED WIRELESS CHANNELS SUPPORTING QOS SERVICES PATRICK HOSEIN Ericsson Wireless Communications, Inc., 6455 Lusk Blvd, San Diego, CA 92121, USA E-mail: patrick.hosein@ericsson.com Third Generation (3G) wireless networks all support a shared high-speed downlink channel for efficient transport of data. These channels use several common techniques (fast link rate adaptation, Hybrid ARQ, fast cell selection and fast scheduling) to improve performance. In addition, future 3G networks must provide the necessary framework to allow operators to offer Quality of Service (QoS) based applications. Service differentiation will also become necessary so that operators can provide differing service levels based on user classes. Pricing will be based on these user classes so that the operators can optimize their revenue. Typically, the shared forward link channel will be the bottleneck resource and hence usage of this resource must be optimally managed. This means that admission controls must be used to ensure that admitted users receive their QoS guarantees for the duration of their connection. If Class-Based Grade of Service (GoS) is offered then the admission control algorithm must also provide class dependent connection blocking rates. In this paper we present a framework that allows an operator to provide Class-Based connection admissions. 1 Introduction Third Generation wireless networks all support some form of a shared high-speed downlink channel for ef- ficient transport of data (the Forward Packet Data Channel or F-PDCH in the case of 1X-EVDV and HRPD networks and the High Speed Downlink Shared Channel or HS-DSCH in the case of WCDMA networks). These channels all use several common techniques (fast link rate adaptation, Hybrid ARQ, fast cell selection and fast scheduling) to improve the throughput and delay performance for data applica- tions. In addition, these networks must provide a flexible framework to allow operators to offer QoS based services (e.g., streaming, gaming, Voice over IP). Service differentiation will also become neces- sary so that operators can provide differentiated lev- els of service based on the subscription plan of the user and seek to maximize revenue. This all implies that a flexible framework is needed to allow for so- phisticated pricing plans. In practice, the shared forward link channel will be the bottleneck resource. This must therefore be carefully managed through (a) optimal scheduling to ensure that users attain their subscribed service levels and the corresponding QoS level guarantees, (b) optimal congestion control to ensure that sys- tem performance does not degrade under congestion and that during these overload periods resources are distributed based on class levels and (c) optimal ad- mission control to ensure that sufficient resources are reserved for already admitted users so that their QoS guarantees can be maintained and also ensure that blocking and dropping rates are class dependent. In a wireless environment, the stochastic capacity avail- able to users together with the bursty nature of the user applications complicates the scheduling, conges- tion control and admission control algorithms. These three resource management algorithms must be jointly designed for optimal system perfor- mance. In this paper we present an admission control algorithm for a generic shared high-speed forward packet data channel. This algorithm tries to main- tain the QoS guarantees of the admitted users and also provide connection blocking and dropping rates based on user classes. Since the admission control algorithm depends on the optimization framework used by the scheduler, we first provide an overview of the corresponding scheduler. We then present the admission control algorithm together with a simple illustrative example. Although congestion controls are not addressed in this paper it is important that such controls also be implemented. Many papers have been written on call admis- sion control algorithms for both TDMA and CDMA based wireless networks where a constant rate, ded- icated channel is provided to the user. In such cases the management of resources for QoS services is well understood since it is quite similar to the wired net- work environment (e.g., see Choi 1 , Chou 2 , Hou 3 , Iraqi 4 and Kwon 5 ). However, providing QoS ser- vices over a shared channel with stochastic capacity is quite more challenging.