基于分层排队网络模型的MCU性能预测及优化研究

针对视频会议系统中的多点控制单元(MCU)提出了一个基于分层排队网络模型的性能预测方法．通过对分层排队网络模型进行扩展，建立了多点控制单元的分层排队模型．设计了一个仿真程序对模型进行评价，仿真程序的输入即模型配置文件．该方法可以较快地从多种设计方案或硬件平台中选择满足系统性能设计目标的软件或硬件配置．仿真结果显示，选择双处理器配置和采用多线程技术，多点控制单元的接入容量提高了50％．最后，多点控制单元终端接入实验证明了该方法的有效性。

LQN Based Performance Model for MCU Performance Prediction

Performance models are powerful tools to evaluate potential system configuration. The layered queue network (LQN) model is widespread used to predict performance of parallel and distributed system. In this paper LQN model is extended by import pseu task and operation task to build performance model of H.323 multipoint control unit (MCU) which will be performance bottle neck in large scale videoconference system. The model build by analyzing media data flow throughout MCU software and hardware include OS and network diver. The model parameters are obtained by several methods such as using network measure tool, adding check point in software, analyzing network data flow. Then the performance parameters can be easily achieved using simulation method. Different software and hardware architecture are evaluated using the LQN model with only different model parameters. By this method it is found the terminal connection capability of MCU can be improved 50% using dual CPUs and multithread architecture in LQN model simulation. The method is proved valid by comparing results with a real system.