| 一种动态优先级排序的虚拟机I/O调度算法 |
| |
| 引用本文： | 郭松辉,龚雪容,王炜,李清宝,孙磊.一种动态优先级排序的虚拟机I/O调度算法[J].计算机科学,2017,44(1):13-19. |
| |
| 作者姓名： | 郭松辉 龚雪容 王炜 李清宝 孙磊 |
| |
| 作者单位： | 解放军信息工程大学 郑州450001;数学工程与先进计算国家重点实验室 郑州450001,解放军信息工程大学 郑州450001,解放军信息工程大学 郑州450001;数学工程与先进计算国家重点实验室 郑州450001,解放军信息工程大学 郑州450001;数学工程与先进计算国家重点实验室 郑州450001,解放军信息工程大学 郑州450001 |
| |
| 基金项目： | 本文受国家自然科学基金(61072047),国防预研基金(910A26010306JB5201)资助 |
| |
| 摘 要： | I/O任务调度是影响I/O密集型虚拟机性能的重要因素。现有调度方法主要是针对虚拟机整机I/O带宽的优化,较少兼顾各虚拟域与全局性能,也无法满足域间差异化服务的要求。针对现有方法的不足,提出了一种动态优先级排序的虚拟机I/O调度算法DPS。该算法基于多属性决策理论,以离差最大化方法计算I/O任务的优先级评估属性权重,对I/O任务优先级进行综合评估；通过引入任务所在虚拟域价值,体现云计算环境下虚拟域重要性差异。在Xen系统中通过实验评测DPS调度虚拟化网卡的性能,结果表明,DPS能够有效提高指定域与全局的I/O任务截止期保证率、整机I/O带宽,并能为不同虚拟域的I/O应用提供差异化服务。
|
| 关 键 词： | 云计算 虚拟化 I/O调度 差异化服务 多属性决策 |
| 收稿时间： | 2015/12/19 0:00:00 |
| 修稿时间： | 2016/3/7 0:00:00 |
| 作者简介： | 郭松辉(1979－),男,博士生,主要研究方向为云计算、虚拟化、性能建模,E-mail:guo_song_hui@163.com;龚雪容(1975－),女,博士,讲师,主要研究方向为网络与服务计算、虚拟化;王 炜(1975－),男,博士,副教授,主要研究方向为云计算可靠性、虚拟化、系统结构;李清宝(1967－),男,博士,教授,博士生导师,主要研究方向为云计算、系统结构;孙 磊(1973－),男,博士,副研究员,硕士生导师,主要研究方向为云计算安全、虚拟化。 |
| |
I/O Scheduling Algorithm Based on Dynamic Prioritization in Virtual Machines |
| |
| GUO Song-hui,GONG Xue-rong,WANG Wei,LI Qing-bao and SUN Lei.I/O Scheduling Algorithm Based on Dynamic Prioritization in Virtual Machines[J].Computer Science,2017,44(1):13-19. |
| |
| Authors: | GUO Song-hui GONG Xue-rong WANG Wei LI Qing-bao SUN Lei |
| |
| Affiliation: | PLA Information Engineering University,Zhengzhou 450001,China;State Key Laboratory of Mathematical Engineering and Advanced Computing,Zhengzhou 450001,China,PLA Information Engineering University,Zhengzhou 450001,China,PLA Information Engineering University,Zhengzhou 450001,China;State Key Laboratory of Mathematical Engineering and Advanced Computing,Zhengzhou 450001,China,PLA Information Engineering University,Zhengzhou 450001,China;State Key Laboratory of Mathematical Engineering and Advanced Computing,Zhengzhou 450001,China and PLA Information Engineering University,Zhengzhou 450001,China |
| |
| Abstract: | I/O scheduling is a key factor for the performance of I/O-intensive virtual machine.The existing scheduling methods mainly focus on optimizing the global I/O bandwidth of virtual machine,and few taking the performance of each virtual domain and the entire system synchronously into account.According to the deficiencies of the existing methods,I/O scheduling algorithm DPS was proposed based on dynamic prioritization.The algorithm takes advantage of multi-attribute decision making theory to calculate attribute weights dynamically for assessing I/O task priority comprehensively,and effectively characterize virtual domain by its value.The performance of DPS is evaluated by scheduling virtualized NIC in Xen.The results show that DPS improves the deadline guarantee ratios of specified domain and the entire system,improves global I/O bandwidth,and it can provide differentiated services for each virtual domain on demand. |
| |
| Keywords: | Cloud computing Virtualization I/O scheduling Differentiated service Multi-attribute decision making |
|
|
|
|
下载PDF全文