Performance Analysis of the ATM Adaptation Layer 2 (AAL2)

In this paper, we analyze the performance of AAL2 multiplexer for a continuous time Markovian arrival process. AAL2 CPS (Common Part Sublayer) packets are multiplexed in the AAL2 multiplexing queue and transmitted in the transmission queue. This tandem structure suggests that the statistics of AAL2 CPS requires at least 2 dimensional state space. Furthermore, from a network-level point of view, cell multiplexing and de-multiplexing procedures are repeated at each AAL2 switching node. That requires simple analysis model. To solve this problem, we reduce the state space by showing that the output process of multiplexing queue can be modeled with the Coxian distribution. We propose a single dimension analysis model of the CPS transmission queue. When AAL2 convey both real and non real time short packets, QoS management is a problem. This is because the QoS of real time as well as non-real time packets is measured using different metrics – delay and cell loss ratio respectively. Most previous work is concentrated around delay performance due to the real time applications getting the primary attention. From the direct comparison of delay and CLR performance, we show that delay constraint is the dominant parameter in QoS of AAL2.