云存储任务调度策略研究
发布时间：2019-02-16 16:21
【摘要】：随着云存储在数据存储方面重要性不断提升,云存储任务调度策略已经成为云存储中一个很重要的方面。云存储任务调度策略主要研究云存储任务分发、完成的过程,随着科技的发展,云存储任务调度策略已经成为云存储中一个关键部分。云存储任务调度策略主要解决的问题是云存储系统怎样选择更优任务调度策略来完成用户提交的任务。本文通过对3个主要的分布式文件系统架构GFS、HDFS、Dynamo及其任务调度策略研究,总结出9个云存储任务调度策略的主要影响因素,副本覆盖范围指标、维护副本一致性代价、副本的可用性、虚拟节点负载、虚拟节点开销、网络传输可靠性、数据副本一致性、可用网络带宽、网络传输延迟。根据对于9个主要影响因素不同的重视程度将用户群划分为4种,分别为个人用户、成长型企业用户、一般规模企业用户、金融类企业用户。本文从4种用户的角度出发,通过与AHP层次分析结合确定云存储任务调度策略中9个因素对于这4种用户各自的权值。在此基础上提出了基于用户角度的云存储任务调度策略。最后,本文将分布式系统的模拟软件optorsim模块进行扩展后,对于所提出的策略进行模拟,综合模拟出的数据得出结论：1)金融类企业用户考虑的因素较多,因此在整个决策中占权值较多。2)成长型企业用户和一般规模企业用户在决策中所占权值相当。3)个人用户考虑因素较少,因此在整个决策中占权值较少。4)在基于4种用户角度的基础上,结合云存储任务调度策略影响因素进行模拟,得出各个节点适合的用户。
[Abstract]:With the increasing importance of cloud storage in data storage, cloud storage task scheduling strategy has become a very important aspect of cloud storage. The task scheduling strategy of cloud storage mainly studies the process of task distribution and completion of cloud storage. With the development of science and technology, the task scheduling strategy of cloud storage has become a key part of cloud storage. The main problem of cloud storage task scheduling strategy is how to select a better task scheduling strategy to complete the tasks submitted by users. Based on the research of three main distributed file system architecture (GFS,HDFS,Dynamo) and its task scheduling strategy, this paper summarizes nine main influencing factors of cloud storage task scheduling policy, replica coverage index, cost of maintaining replica consistency. Replica availability, virtual node load, virtual node overhead, network transmission reliability, data replica consistency, available network bandwidth, network transmission latency. According to the different degree of attention to the nine main influencing factors, the user group is divided into four types: individual user, growth enterprise user, general scale enterprise user and financial enterprise user. From the point of view of four users, this paper determines the respective weights of 9 factors in the cloud storage task scheduling policy by combining with the AHP hierarchy analysis. On this basis, a cloud storage task scheduling strategy based on user's perspective is proposed. Finally, after extending the simulation software optorsim module of the distributed system, this paper simulates the proposed strategy, synthesizes the data obtained from the simulation and draws a conclusion: 1) the financial enterprise users consider many factors. Therefore, in the whole decision making, the weight value is much higher. 2) the growth enterprise user and the general scale enterprise user share the same weight value in the decision making. 3) the individual user considers the factor less. Therefore in the whole decision-making less weight. 4) on the basis of four user perspectives combined with cloud storage task scheduling policy factors to simulate and get the appropriate users for each node.
【学位授予单位】：西安工业大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP333
本文编号：2424627
[Abstract]:With the increasing importance of cloud storage in data storage, cloud storage task scheduling strategy has become a very important aspect of cloud storage. The task scheduling strategy of cloud storage mainly studies the process of task distribution and completion of cloud storage. With the development of science and technology, the task scheduling strategy of cloud storage has become a key part of cloud storage. The main problem of cloud storage task scheduling strategy is how to select a better task scheduling strategy to complete the tasks submitted by users. Based on the research of three main distributed file system architecture (GFS,HDFS,Dynamo) and its task scheduling strategy, this paper summarizes nine main influencing factors of cloud storage task scheduling policy, replica coverage index, cost of maintaining replica consistency. Replica availability, virtual node load, virtual node overhead, network transmission reliability, data replica consistency, available network bandwidth, network transmission latency. According to the different degree of attention to the nine main influencing factors, the user group is divided into four types: individual user, growth enterprise user, general scale enterprise user and financial enterprise user. From the point of view of four users, this paper determines the respective weights of 9 factors in the cloud storage task scheduling policy by combining with the AHP hierarchy analysis. On this basis, a cloud storage task scheduling strategy based on user's perspective is proposed. Finally, after extending the simulation software optorsim module of the distributed system, this paper simulates the proposed strategy, synthesizes the data obtained from the simulation and draws a conclusion: 1) the financial enterprise users consider many factors. Therefore, in the whole decision making, the weight value is much higher. 2) the growth enterprise user and the general scale enterprise user share the same weight value in the decision making. 3) the individual user considers the factor less. Therefore in the whole decision-making less weight. 4) on the basis of four user perspectives combined with cloud storage task scheduling policy factors to simulate and get the appropriate users for each node.
【学位授予单位】：西安工业大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP333
【参考文献】
相关期刊论文 前2条
1 卢正鼎;双机系统上的一个作业调度算法[J];华中工学院学报;1985年06期
2 王永海;;2009存储技术热点之IP存储和存储虚拟化[J];科技浪潮;2009年02期
，本文编号：2424627
本文链接：https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/2424627.html
