一种面向软件仓库挖掘的动态作业配置框架

Research on Dynamic Job Configuration Framework for Mining of Software Repositories

构造面向软件仓库挖掘的数据中心,是目前软件工程领域的研究热点。软件仓库数据处理作业的执行时间差异明显、资源消耗大等特点为其作业配置带来诸多挑战。提出一种面向软件仓库挖掘的作业配置框架TrustieS-DC,该框架支持一种新型远程作业部署和服务模式,采用一种基于软件版本划分的动态作业配置算法以缩短长作业响应时间并提高系统资源利用率。基于Gnome项目SVN库的实验表明,TrustieSDC的性能和资源利用率与并行后的Alitheia相比有明显改进。

Construction of datacenters for mining of software repositories（MSR） is a hot topic in current software engineering area.Data processing jobs for software repositories are highly diverse in execution-time and resource-consuming,which bring many challenges for the job configuration in such environments.This paper proposed a job configuration framework named TrustieSDC for mining of software repositories.TrustieSDC supports a new paradigm for remote deployment and execution of MSR jobs,and proposes a software-subversion-partition based job configuration algorithm to cut the response time of long jobs and increase the resource utilization.The experiments based on SVN repositories of Gnome projects shows that compared with paralleled Alitheia system,TrustieSDC gains remarkable improvement on both performance and resource utilization.