基于统计预测的Cloudlet调度机制研究  

Cloudlet Scheduling Mechanism Research Based on the Statistical Forecasting

作　　者：华夏进 董瑞志[1,2] 彭鑫[1,2] 赵文耘[1,2] 

机构地区：[1]复旦大学软件学院,上海201203 [2]上海市数据科学重点实验室(复旦大学),上海201203

出　　处：《小型微型计算机系统》2016年第3期406-411,共6页Journal of Chinese Computer Systems

基　　金：国家"八六三"高技术研究发展计划项目(2013AA01A605)资助

摘　　要：Cloudlet利用虚拟机来灵活处理来自不同移动设备的云应用服务访问请求,避免了移动设备直接访问云应用服务过程中的广域网延迟和带宽限制问题.但是基于虚拟机实现的Cloudlet,存在合成虚拟机耗时长的问题,对于用户体验影响很大.针对这个问题,本文提出一种基于统计预测的调度机制,该机制使用了两种预测模型,分别是基于Cloudlet切换统计模型和基于关键位置的统计预测模型,来预测移动中的用户将会使用的Cloudlet.然后,根据预测信息实现Cloudlet上应用服务虚拟机预先调度合成,以降低用户请求等待时间.实验表明,与其它方法相比,基于本文预测机制的方法可以有效降低用户的请求等待时间,提高用户体验.Cloudlet use the flexibility of the virtual machine to handle requests from different mobile devices,which avoids the problem of WAN latency and bandwidth limitation in the case of mobile devices access cloud service directly. But it is time-consuming for Cloudlets to synthesize virtual machines,which has significant impact on user experience. To address such issue,we propose a statistical-based forecasting approach to guide the scheduling of the Cloudlet resources. The method encapsulates two kinds of statistical model,which are statistical model based on the Cloudlet switching and statistical model based on the key positions,to predict the Cloudlet which the user will access. Then,our method will guide the selected Cloudlet on howto synthesize virtual machine in advance so as to reduce user waiting time. Experiment results showthat our method can reduce user waiting time effectively,and improve the user experience considerably.

关 键 词：预测调度 Cloudlet调度 移动计算 资源限制 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]