For the N * D/D/1 queue, the exact formulas for the virtual waiting time distribution and the waiting time distribution are well known. The exact formulas are extended to the case of the /spl Sigma/ N/sub k/D(X/sub k/)/D/1 queue, where different arrival processes can have different packet sizes. Numerical examples are also presented to illustrate their practical applications.

While the efficiency of call admission control (CAC) schemes can be improved by taking advantage of call-level behavior, it is also important to make sure that the reliability of the CAC schemes is maintained. This paper studies the reliability aspect of CAC schemes for the reference model that consists of the M/M/c loss system at the call level, and the bufferless link with single-class exponential on-off fluid traffic at the burst level. As a measure of reliability, the probability that an admitted call violates a loss-ratio target is used, and its analytic expression is derived for the reference model by using performability analysis and the nearly completely decomposability approximation. It is shown that the CAC schemes previously proposed to take advantage of call-level behavior are not acceptably reliable. In addition, it is shown that, for most cases in practice, if the reliability of a CAC scheme taking advantage of call-level behavior is to be maintained at a reasonable level, the improvement in efficiency relative to the static scheme will be small.

The paper addresses the following issues related to the performance and design of the ATM/AAL2 transport in UTRAN (UMTS terrestrial radio access network): 1) it proposes an alternative AAL2 implementation method that has advantages over the widely-used method; 2) it studies the effect of the stringent delay requirement on the bandwidth requirement of the ATM/AAL2 transport; 3) it proposes the use of common part sublayer (CPS) packet shaping to deal with the burstiness of CBR-based dedicated channel (DCH) traffic caused by the periodic nature of the MAC layer of the UMTS radio interface; 4) it evaluates statistical multiplexing gain from the ATM/AAL2 transport as compared with TDM transport.