Simulation Modeling And Performance Analyses Of ATM Switches

The broadband integrated services digital network ( B_ISDN ) with standardized asynchronous transfer mode (ATM) is envisaged to support not only current serices, but also new services with varying traffic characteristics and quality-of-service (QoS) requirements.So it is widely used in high speed bone networks due to its efficiency and flexibility~[1-4] The exchange architectures of ATM switches which are responsible for buffering and putting foword ATM cells, are the key tache of the ATM technology and have a directly effect on the performance and quality of the network. Buffering strategies or queueing strategies, which are the important part of the ATM exchange architectures' design, have a great effect on the performance of transmition of the packets such as delay and loss. So they still attract widely attentions and researches. For the complicacy of modeling the buffering strategies, theory deriving cannot fulfil varying business models and buffering strategies. So the system simulation is adopted more often.The system simulation is said to be the third measure of understanding and changing world, following the fundamental research and the experimental technology. After studying the basic theory of the queueing models and the system simulation, we applied it in the buffering strategies of ATM network switches and did a deep research. Based on the principle of the simulation of the discrete event system, this article proposed the simulation models of cell buffering system of the ATM switches is proposed in this paper, and computer simulation is used to analyse and compare the different buffering strategies and business models.First, based on the principle of the simulation of the discrete event system, the simulation models of three basic output buffering management schemes were proposed, which are complete partitioning, complete sharing and partial sharing. And the simulation modules of cell arriving and cell departing were present. And the simulation program was run under the even business model and bursty business model, and came to a conclusion that the simulating data were in agreement with the theory data and the analyse taillied with the actual situation. These simulation models for the ATM switches laid a foundation of further deep research.Afterwards, the modeling and the analysis in view of the dynamic threshold strategy are carried on. Among the several buffering strategies the output complete sharing buffer strategy iswidely used due to its efficiency and moderate complexity. In 1998 Abhijit K. Choudhury and Ellen L.Hahne proposed the dynamic threshold concept. The philosophy behind it is adaptation of the threshold for each output logical queue to the queue length of the cells. A lot of literatures based on the dynamic threshold strategy are proposed in recent years~[5,6] . Based on the modeling and analyse on the original dynamic threshold scheme, this article proposed an adaptive dynamic threshold scheme which is adaptation of the threshold to the average queue length. Through the simulation, it could be found that our scheme is more adaptive and has a lower cell loss percentage.Lastly, the modeling and the analysis in view of the adaptive priority services is carried on. Since today broadband communication service requests a wider bandwidth and a higher speed, a lot of mix businesses appear such as multimedia communication. As the exchanges require a better quality-of-service, one of the effective ways is concerning about priority of the cells in the switches. So this article extended our adaptive dynamic threshold shceme to the priority service models and established its simulation modeling. Through the analyses and comparisons of the simulation data, it could be found that our scheme has a better performance in the priority service too.Simulation modelings of buffer system of the ATM switches and implemented all buffering strategies discussed in the paper has been established. Through computer simulation, it is confirmed that the buffering strategies our proposed have a better performance. This...