一种基于关联挖掘的服务一致化配置方法  

Association Mining Based Consistent Service Configuration

作　　者：王焘[1,2] 陈伟[2] 李娟[3] 刘绍华[4] 苏林刚 张文博[1,2] Wang Tao;Chen Wei;Li Juan;Liu Shaohua;Su Lingang;Zhang Wenbo(State Key Laboratory of Computer Science(Institute of Software,Chinese Academy of Sciences),Beijing 100190;Institute of Software,Chinese Academy of Sciences,Beijing 100190;Beijing University of Technology,Beijing 100124;Beijing University of Posts and Telecommunications,Beijing 100876)

机构地区：[1]计算机科学国家重点实验室(中国科学院软件研究所),北京100190 [2]中国科学院软件研究所,北京100190 [3]北京工业大学,北京100124 [4]北京邮电大学,北京100876

出　　处：《计算机研究与发展》2020年第1期188-201,共14页Journal of Computer Research and Development

基　　金：国家重点研发计划项目(2017YFB1400804);国家自然科学基金项目(61872344);北京市自然科学基金项目(4182070);中国科学院青年促进会人才专项项目(2018144)~~

摘　　要：组件化服务化软件系统由松耦合的异构服务组件构成,每个服务组件都包含着大量可高度灵活配置的配置项.服务组件之间存在着复杂的依赖关系,导致其配置项相互关联,使得系统部署、更新或迁移易于出错.对于相互关联的配置项,更改一个配置项就需要修改与之关联的其他配置项,否则将违反约束条件,导致系统出现故障.因而,分析配置项关联性对于保障系统可靠性至关重要,但需要跨产品的领域知识.提出了一种基于关联挖掘的服务一致化配置方法.该方法爬取配置文件样本数据以将搜索范围缩小到频繁改变的配置项,根据配置项的名称、取值和类型的相似性计算,为配置项对生成关联系数,使用定义的过滤规则确定候选关联配置项对集合,输出排序的配置项关联性列表以供查询.基于该方法部署了典型应用系统进行实验和评估,实验结果表明:该方法能够准确检测配置项的关联性.Componentized service-oriented software systems always consist of loosely coupled hetero-geneous service components,each of which contains a large number of configuration items configured with high flexibility.Complex dependencies exist between service components,resulting in their interrelated configuration items,so the operations of deploying,updating and migrating components are prone to errors.For configuration items related with each other,changing one configuration item requires to modify other related configuration items.Otherwise,violating constraints perhaps happens,which should cause system failure.Therefore,analyzing associations between configuration items is one key to ensure the reliability of a system.This paper proposes a service configuration approach based on association mining.Our approach crawls configuration files from Internet,narrows the analysis scope to frequently changed configuration items,generates association coefficients for item pairs according to the similarity of items name,value and type,determines the set of candidate item pairs with rules,and then outputs a list of ordered configuration item pairs for query.We have deployed two typical open-source software systems to validate our approach for mining configuration associations between configuration items with case studies.Experimental results show that our approach can accurately detect most associations between configuration items.

关 键 词：关联挖掘 服务配置 配置关联 错误检测 服务可靠性 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]