Cross-Layer Connection Admission Control Policies for Packetized Systems

Abstract

Delivering quality of service in packetized mobile cellular systems is costly, yet critical. Recently, cross-layer connection admission control policies [1] [2] have been shown to realize network performance objectives for multimedia transmission that include constraints on delay and blocking probability. Current third generation (3G) systems such as high speed uplink packet access (HSUPA) employ a threshold-based admission control (AC) policy to reserve capacity to increase quality of service (QoS). In threshold-based AC, a user request is admitted if the load reported is below a threshold. Although a threshold-based AC policy is simple to implement and may be improved upon to take into account resource allocation information [3], it unfortunately cannot meet upper layer QoS requirements, such as required in the data-link and network layers [4]. In this chapter, AC policies are investigated for packetized code division multiple access (CDMA) systems that can both maximize overall system throughput and simultaneously guarantee quality of service (QoS) requirements in both physical and upper layers. To further improve user capacity, multiple antennas are employed at the base station, and a truncated automatic repeat request (ARQ) scheme is employed in the data link layer of the system under investigation. Truncated ARQ is an error-control protocol which retransmits an erroneous packet until either it is correctly received or until a maximum number of retransmissions is reached. The design of optimal connection admission control policies for a packetized CDMA system that incorporates an advanced multi-beamformer basestation at the physical layer and ARQ at the data link layer has, to the authors’ knowledge, not been addressed previously. For example, the call level admission control policies for CDMA systems in [4] [5] [6] only focus on circuit-switched networks, in which radio resources allocated to a user are unchanged throughout the call connection, leading to inefficient utilization of system resources, especially for bursty multimedia traffic. In [7] [8], the CAC problem is extended to packetswitched CDMA systems. Unfortunately, the CAC modelling in [7] [8] has been limited to optimizing power control and admission control policies to specific systems, in which physical layer performance, characterized in terms of signal-to-interference (SIR) in each service class, is static. With multiple antennas systems, which are widely employed in current 3G CDMA systems [9] [14], the physical layer performance depends not only on system state, but also on factors such as spatial angle of arrival (AoA). Therefore, the existing CAC framework in [7] [8] cannot adequately incorporate multiple antenna base-