支持软件按需流式加载的预取机制  

Prefetching Mechanism for On-demand Software Streaming

作　　者：钟亮[1] 胡春明[1] 沃天宇[1] 李建欣[1] 康俊彬[1] 

机构地区：[1]北京航空航天大学计算机学院,北京100191

出　　处：《计算机研究与发展》2011年第7期1178-1189,共12页Journal of Computer Research and Development

基　　金：国家"九七三"重点基础研究计划基金项目(2011CB302602);国家自然科学基金项目(60703056)

摘　　要：近年来,随着SaaS技术的发展,软件的网络化、服务化访问成为一种新的使用模式.软件的按需动态部署是实现上述模式的重要基础.为了支持软件的按需动态部署,需要能够在执行环境支持软件的流式加载运行.而在软件按需流式加载的执行过程中,程序会因为请求缺失的数据块被阻塞直至数据块被下载过来,从而极大地影响执行性能与用户体验.针对流式加载中的性能问题,提出一种基于N-Gram预测模型和增量数据挖掘技术的预取机制,该预取机制可用于支持软件流式加载执行.预取机制通过收集用户使用软件所产生的历史访问日志,进行数据挖掘分析,来动态更新、完善预取规则,然后根据最合理的预取规则进行软件预取.该预取机制可同时支持基于文件级别和软件块级别的预取.实验结果表明,对于各类软件,该可预取的文件系统能够将软件启动加载时间减少10%～50%,而预取命中率达到了81%～97%.In recent years, the “software as a service”, largely enabled by the lnternet, has become an innovative software delivery model to provide network and service accessing of software. Dynamic on demand deployment of software is a key method to achieve the above delivery model. Software streaming delivering is needed to support this deployment manner. During the streaming delivering of software, the execution waits until the missing data block is downloaded, which greatly influences the execution performance and user experience. A prefetching mechanism is presented for software streaming delivering based on N-Gram prediction model and an incremental data mining algorithm, By using historical access logs for data mining, then dynamically updating and polishing the prefetching rules, the proposed prefetehing framework supports both file-level prefetching and block level prefetching. The experimental results show that this prefetch-enable filesystem achieves a launch time reduced by 10% to 50%, as well as hit rate between 81% and 97%.

关 键 词：软件即服务 流式加载 预取 日志挖掘 虚拟计算 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]