【作者】 姜云汉；

【导师】 李志淮； 刘朝斌；

【作者基本信息】 大连海事大学 ， 计算机应用技术， 2007， 硕士

【摘要】 网络业务流的自相似性对于网络链路传输控制过程中的丢包率、吞吐量以及队列容量会产生严重的影响，导致自相似业务流下的网络链路传输控制性能与传统模型的结论有较大差异，意味着传统的基于马尔可夫模型的性能分析结果对自相似业务已不再适用，需要采用新的模型方法来对自相似业务下的网络链路传输控制性能进行性能分析。 本文深入分析网络链路传输控制中不同类型业务流的特点，并建立不同类型业务流的随机Petri网模型，同时针对自相似业务流的影响，提出网络链路传输控制中的具有多种传输源的随机Petri网模型。然后尝试采用一种非马尔可夫随机Petri网的方法对该模型进行分析，最后利用工具SPNP对所分析内容进行了模拟和验证。 实验结果表明：在自相似业务流的影响下，网络链路传输控制的吞吐量、丢失率和重传率也平缓地增长，而平均队列容量快速增加，具有较高的突发性，因此需要更多的带宽。这样为提高网络链路传输控制性能提供了定量分析方法，并为网络链路传输控制性能分析提供了非马尔可夫模型。更多还原

【Abstract】 Self-similarity nature of network traffic has severe effect on packet loss rate, throughput and queue capacity of network link transfer control. This leads to significant results difference between the traditional model and self-similarity traffic in network performance. Therefore, the traditional Markovian model is no longer applicable to analyze performance of self-similarity traffic. There is a great need to adopt new models and tools to study the performance of network link transfer control with self-similarity nature.Based on deep analysis of the characteristics of different traffic flow in network link transfer control, stochastic Petri net models for different traffic flow are presented. As far as the effects of self-similarity on network communication are concerned, diversified transfer source stochastic Petri net models are put forward in this paper. Then, a non markovian stochastic Petri net method is adopted to analyze the proposed models with simulation and verification based on SPNP tool.Simulation results show that throughput, packet loss rate and retransmission rate of network transfer control grows slowly for self-similarity traffic flow, but the average queue capacity increases quickly characterized with bursts, therefore more bandwidth is needed. Our work provides a quantitative analysis approach for improving the performance of network link control in different conditions and presents a non markvian model to analyze the performance.更多还原