The Research On Wireless System Traffic Model And Call Blocking Performance Improvement Strategy

With the internet service demand increase sharply in recent years, the Congestion Control, Bandwidth Allocation, Call Admission Control(CAC) in high-speed network have become the hotspot of network research. In communication domain, the descrip- tion of the high-speed network's characteristic in self-similar traffic model is more truthful than that in traditional traffic model, so the research in simulation of self- similar traffic model is very effective. Based on systemic introduction of the random process theory, the paper investigated the call blocking performance of self-similar traffic deeply. As the M/M/C/K model with Poisson arrival and exponential distribut- ion serve time, the paper derived the CBP formula when channels and buffers both be taken up. For M/P/C/K model with Pareto distribution serve time, we analyzed it detailedly, and got the two model's law of CBP by the change of buffer and load. The experimental results show that the self-similar traffic results in call blocking perfor- mance deterioration.In mobile communication system, CAC can control the number of user talking synchronously in network directly, avoiding network Congestion, decreasing call lost,so it can improve the quality of service. Thus, it should be studied widely and deeply as one of wireless resource management scheme. At present, there exists two kinds of CAC scheme: guard channels and queue priority. In this paper, we consider the guard channels method. To realize Real-time Call Admission Control(RCAC), it's very important to get the number of guard channels quickly and accurately,when such parameters as the handover call dropping probability, the number of cell channels and both new and handover calls'load are specified. This paper derived an approximate formula of guard channels and then presented a new algorithm. We validates that, the number of guard channels got by new algorithm can give high precision, but the running time reduces dramatically, when compared to the multi- iteration probing algorithm.