可演化网络中移动代理的性能分析及优化方法  

Performance Analysis and Optimization for Mobile Agents in Evolutionary Network

作　　者：刘涛[1,2] 钱德沛[1,3] 黄泳翔[1] 王锐[3] 栾钟治[3] 

机构地区：[1]西安交通大学电子与信息工程学院,西安710049 [2]中国科学院深圳先进技术研究院,广东深圳518055 [3]北京航空航天大学计算机学院,北京100191

出　　处：《西安交通大学学报》2010年第12期5-9,70,共6页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(90104022);国家"863计划"资助项目(2006AA01A109;2009AA01Z144)

摘　　要：针对以移动代理软件方式实现的可演化网络的性能问题,提出了一种快速激活及加载方法.通过对移动代理非执行转发、执行转发及执行各步骤的详细性能测量及分析,发现网络节点的执行环境为移动代理执行所做的准备工作时间消耗较大.该快速方法首先解开移动代理包中的所有程序代码并进行预先加载,在加载主程序时优先搜索已预先加载的代码,从而能够较快地找到程序代码.快速方法通过优化代码加载时的搜索顺序,有效地加快了可演化网络中移动代理的执行过程.测试结果表明,采用快速方法使典型的移动代理平均启动加载时间缩短了3.26 ms,网络节点的吞吐率提高了58%.A novel fast launch algorithm is proposed to improve the performance of mobile agents in evolutionary networks.Analyses on the performance data of mobile agent＇s non-execution forwarding,execution forwarding,and detail execution steps show that the execution environment preparing process for mobile agents is the bottleneck.When the fast algorithm is executed,all codes in the mobile agent package will be extracted and loaded prior to the mobile agent＇s main program.Before a new code is loaded,the preloaded codes will be searched,and hence the code image can be located quickly.The mobile agent＇s execution forwarding can be improved by optimizing the searching order.Testing results show that the average loading time of the mobile agents is reduced by 3.26 ms and the throughput of network nodes is increased by 58%.

关 键 词：可演化网络 移动代理 性能优化 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]