Performance Analysis for Multimedia Communication Systems with a Multilayer Queuing Network Model  

作　　者：Xuehua Tang Zhongyuan Wang Xiaojun Li Zhen Han Zheng He Youming Fu 

机构地区：[1]School of Remote Sensing and Information Engineering,Wuhan University,Wuhan 430079,China [2]Research Institute of Wuhan University in Shenzhen,Shenzhen 518000,China [3]School of Computer Science,Wuhan University,Wuhan 430072,China [4]State Grid Economic and Technological Research Institute CO.,LTD,Beijing 102209,China

出　　处：《China Communications》2018年第8期67-76,共10页中国通信（英文版）

基　　金：supported by the Application Research of the Remote Sensing Technology on Global Energy Internet(JYYKJXM(2017)011);the National Natural Science Foundation of China(61671332,41701518,41771452,41771454,U1736206);National key R&D Project(2016YFE0202300);Hubei Province Technological Innovation Major Project(2017AAA123);Applied Basic Research Program of Wuhan City(2016010101010025);Basic Research Program of Shenzhen(JCYJ20170306171431656);the Fundamental Research Funds for the Central Universities(2042016gf0033)

摘　　要：Software performance evaluation in multimedia communication systems is typically formulated into a multi-layered client-server queuing network(MLCSQN) problem. However, the existing analytical methods to MLCSQN model cannot provide satisfactory solution in terms of accuracy, convergence and consideration of interlocking effects. To this end, this paper proposes a heuristic solving method for MLCSQN model to boost the performance prediction of distributed multimedia software systems. The core concept of this method is referred to as the basic model, which can be further decomposed into two sub-models: client sub-model and server sub-model. The client sub-model calculates think time for server sub-model, and the server sub-model calculates waiting time for client sub-model. Using a breadthfirst traversal from leaf nodes to the root node and vice versa, the basic model is then adapted to MLCSQN, with net sub-models iteratively resolved. Similarly, the interlocking problem is effectively addressed with the help of the basic model. This analytical solver enjoys advantages of fast convergence, independence on specific average value analysis(MVA) methods and eliminating interlocking effects.Numerical experimental results on accuracy and computation efficiency verify its superiority over anchors.Software performance evaluation in multimedia communication systems is typically formulated into a multi-layered client-server queuing network （MLCSQN） problem. However, the existing analytical methods to MLCSQN model cannot provide satisfactory solution in terms of accuracy, convergence and consideration of interlock- ing effects. To this end, this paper proposes a heuristic solving method for MLCSQN model to boost the performance prediction of distributed multimedia software systems. The core concept of this method is referred to as the basic model, which can be further decom- posed into two sub-models： client sub-model and server sub-model. The client sub-model calculates think time for server sub-model, and the server sub-model calculates waiting time for client sub-model. Using a breadthfirst traversal from leaf nodes to the root node and vice versa, the basic model is then adapted to MLCSQN, with net sub-models iteratively resolved. Similarly, the interlocking problem is effectively addressed with the help of the basic model. This analytical solver enjoys advantages of fast convergence, independence on specific average value analysis （MVA） methods and eliminating interlocking effects.Numerical experimental results on accuracy and computation efficiency verify its superiority over anchors.

关 键 词：multimedia communication system queuing network performance evaluation 

分 类 号：TN919.8[电子电信—通信与信息系统]