SIP的分析与性能优化

Analysis of SIP and its performance optimization

为了减小会话初始化协议的呼叫建立延时,获得性能和缓冲数量上的平衡点,深入分析SIP复杂的协议行为,并建立协议的广义随机Petri网模型.构造与广义随机Petri网模型同构的马尔可夫链,根据马尔可夫链求解稳定状态概率,对协议模型进行计算和分析,评价协议性能.针对协议运行时状态转换中的耗时过程,通过软件SPNP仿真计算分析得到模型中位置P2缓冲大小的优化结果,得出呼叫建立平均延时随位置P2缓冲大小变化的规律,同时对系统吞吐量进行评价.仿真结果表明,位置P2缓冲大小取优化值时能优于其他取值,缩短协议的呼叫建立延时,改善系统吞吐量,提高系统的性能.

By analyzing thoroughly the complex protocol behavior of SIP, a generalized stochastic Petri network （GSPN） model of protocol was established for reducing call setting up time-delay of session initia- tion protocol （SIP） and obtaining a tradeoff between the performance and the number of buffers. Based on the model, Markov chain was designed, which was isologous with GSPN. Performance index of SIP was studied and evaluated through calculating and analyzing the designed Markov chain. The time-consuming process of SIP＇s service during state transformation was proposed. Aimed at the time-consuming process, the optimal size of the buffer at position Pz was computed by means of simulation with SPNP kit. Mean- while, the simulation was employed to get the variation pattern of average time delay of call setting-up with the buffer size at position Pz, and to evaluate the throughput of system. Simulation result showed that the system performance with optimal buffer size at position P2 was better than that with other buffer size, reducing call setting-up time and improving the performance of protocol.