基于随机搜索规则的软件体系结构层性能演化优化方法  

Random Search Rule Based Performance Evolutionary Optimization Method at Software Architecture Level

作　　者：倪友聪 李松[1] 叶鹏 杜欣 

机构地区：[1]福建师范大学数学与信息学院,福州350117 [2]伦敦大学学院计算机科学学院 [3]武汉纺织大学数学与计算机学院,武汉430200

出　　处：《计算机科学》2017年第11期156-163,174,共9页Computer Science

基　　金：国家自然科学基金(61305079;61370078);福建省自然科学基金(2015J01235;2017501498);福建省教育厅JK类项目(JK2015006);武汉大学软件工程国家重点实验室开放基金(SKLSE 2014-10-02)资助

摘　　要：已有的基于规则的软件体系结构(Software Architecture,SA)层性能优化方法大多未充分考虑优化过程中各规则的改进幅度、使用次数和使用顺序的不确定性,导致了搜索空间受限而难以获取更优的性能改进方案。针对该问题,基于SA层性能改进策略定义一组随机搜索规则,以增大各规则的性能改进空间;在此基础上考虑这些规则的不同使用顺序和不同使用次数的组合情况,构建SA层性能优化模型RRPOM,并设计演化求解算法,进而形成一种SA层性能优化方法 RRMO4PO。与已有方法在WebApp应用案例上的实验对比表明,该方法在使用更少的规则、更少次修改SA元素而获取更好可解释性的同时,有效减少了系统响应时间和改进代价。在最好的情况下,平均使用有改进效果的规则的次数和平均修改SA元素的次数较已有方法分别降低了33.3%和52.9%,与此同时将系统响应时间和改进代价分别降低了30.5%和73.6%。The existing rule-based performance optimization approaches at software architecture（SA）level don＇t fully concern the improvement range of the rule and the uncertainty of the count and the order of each rule usage in the optimization process.As a result,the search space for performance improvement is limited and the better solutions are hard to find.Aiming at this problem,firstly,agroup of the random search rules（RSRs）were designed based on performance improvement tactics so that each RSR can explore the larger performance improvement space.Then the combination of the order and the count of each rule usage was involved in the definition of the random search rule based performance optimization model named RRPOM and the evolutionary algorithm was designed to solve RRPOM.Further,the random search rule based method for performance optimization at software architecture level（RRMO4PO）was proposed.Finally,a WebApp application was taken as a case in the experiments for comparing RRMO4PO with the existing methods.The experimental results show that RRMO4PO can obtain the solutions with better interpretability by using fewer rules and fewer times to modify the SA elements.In addition,the results also prove that both system response time and cost for performance improvement can be decreased more efficiently in our approach.In the best results in our experiments,the average number of rules with improvement effect and the times to modify the SA elements are reduced by 33.3% and 52.9% respectively,and the system response time and the cost for performance improvement are decreased by 30.5% and 73.6% respectively.

关 键 词：软件体系结构 性能优化 系统响应时间 随机搜索规则 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]