Determining the call admission region for real-time heterogeneous applications in wireless TDMA networks

Abstract

The call admission problem for a wireless packet-switched network supporting homogeneous applications-such as a cellular voice network-is an old one and has been extensively investigated in the past. The article investigates the call admission region for a TDMA (wireless) system supporting heterogeneous real-time VBR applications with distinct QoS requirements and traffic characteristics. The QoS is defined in terms of a maximum tolerable packet delay and dropping probability; packets may be dropped due to delay violations or channel-induced errors. The call acceptance region is investigated under the assumption that each user's per-frame resource (slot) requests are communicated to the scheduler (resource allocation authority). The call acceptance region is shaped by the QoS that can be delivered by the uplink scheduling policy. Some of the mechanisms employed to inform the scheduler of the users' requests along with the scheduling policies are discussed. While these scheduling policies identify QoS points in the call admission region, they cannot reveal the entire region. An approach is outlined on how to precisely determine the call admission region (largest set of QoS points delivered under any scheduling policy) for a TDMA system. This approach is shown to lead to the precise description of the reduced call admission region in the presence of channel errors. In order to demonstrate the performance improvement provided by an error control scheme, a simple QoS-sensitive forward error control protocol operating over on erroneous channel is employed, leading to an enlargement of the calculated call admission region. Finally, it is shown that feasible scheduling policies exist which do deliver at least the minimum QoS requirement of the applications whose associated QoS vector falls within the determined call admission region.