云平台下基于可信性的资源调度策略研究
发布时间：2018-10-10 11:58
【摘要】：日前，人们对计算能力、软件服务质量以及大规模数据量的处理要求越来越高，而现有的计算能力不能满足这些需要，于是云计算得以提出。云计算发展到今天，不论是在学术界还是在商业领域都有着非常广泛的应用。科技的发展使得现在数据量的级别从最早的GB级上升到TB级乃至PB级，因此研究出更好的云平台计算服务迫在眉睫。 本文一开始对云计算的定义做了归纳，提到计算能力作为一种商品向用户提供并按使用情况收取服务费用，接着列举了云计算系统的特点以及架构，并对云计算实现的关键技术做了详细的分析，然后介绍了当今流行的云平台。为了能让本文提出的算法在云平台上模拟实验，本文还研究了MapReduce机制的原理、执行流程、 Hadoop的架构等。同时，为了比较本文提出的算法和Hadoop资源调度算法的异同，本文就当今流行的三种作业调度算法：FIFO队列调度算法、Fair公平调度算法以及基于计算性能的Capacity算法做了详细的研究，分析了每一种算法的优劣，以便同本文的算法进行更为详细的比较。 分布在云平台下的节点资源数量非常巨大，这就不可避免的造成了不可靠节点资源的出现，，这些节点会对应用程序的执行和调度任务产生很大的影响。在本文中，受贝叶斯认知模型的启发和社会学的信任关系模型的引导，本文首先提出了一种新的基于贝叶斯方法的认知信任模型，然后，将这种模型应用到资源调度系统中。理论分析和仿真实验证明，本文提出的方法能有效的满足云计算对节点资源的信任要求，并且牺牲较少的时间成本，确保在一个相对安全的节点资源池中执行云计算任务。
[Abstract]:A few days ago, the demands of computing power, software quality of service and large amount of data were higher and higher, but the existing computing power could not meet these needs, so cloud computing was put forward. Cloud computing has been widely used in both academic and commercial fields. With the development of science and technology, the level of data volume has risen from the earliest GB level to the TB level and even the PB level, so it is urgent to develop a better cloud platform computing service. At the beginning of this paper, we summarize the definition of cloud computing, mention that computing power is provided to users as a commodity and charge the service fee according to the usage, and then enumerate the characteristics and architecture of cloud computing system. The key technologies of cloud computing are analyzed in detail, and then the popular cloud platform is introduced. In order to enable the algorithm proposed in this paper to simulate the experiments on the cloud platform, this paper also studies the principle of MapReduce mechanism, execution flow, Hadoop architecture and so on. At the same time, in order to compare the similarities and differences between the proposed algorithm and Hadoop resource scheduling algorithm, this paper makes a detailed study on three popular job scheduling algorithms: FIFO queue scheduling algorithm, Fair fair scheduling algorithm and Capacity algorithm based on computational performance. The advantages and disadvantages of each algorithm are analyzed in order to compare with the algorithm in detail. The number of node resources distributed in the cloud platform is very large, which inevitably leads to the emergence of unreliable node resources, and these nodes will have a great impact on the execution and scheduling of applications. In this paper, inspired by Bayesian cognitive model and guided by sociological trust relationship model, this paper first proposes a new cognitive trust model based on Bayesian method, and then applies this model to resource scheduling system. Theoretical analysis and simulation experiments show that the proposed method can effectively meet the trust requirements of cloud computing to node resources, and at the expense of less time cost, ensure the implementation of cloud computing tasks in a relatively secure node resource pool.
【学位授予单位】：山东师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.01
本文编号：2261683
[Abstract]:A few days ago, the demands of computing power, software quality of service and large amount of data were higher and higher, but the existing computing power could not meet these needs, so cloud computing was put forward. Cloud computing has been widely used in both academic and commercial fields. With the development of science and technology, the level of data volume has risen from the earliest GB level to the TB level and even the PB level, so it is urgent to develop a better cloud platform computing service. At the beginning of this paper, we summarize the definition of cloud computing, mention that computing power is provided to users as a commodity and charge the service fee according to the usage, and then enumerate the characteristics and architecture of cloud computing system. The key technologies of cloud computing are analyzed in detail, and then the popular cloud platform is introduced. In order to enable the algorithm proposed in this paper to simulate the experiments on the cloud platform, this paper also studies the principle of MapReduce mechanism, execution flow, Hadoop architecture and so on. At the same time, in order to compare the similarities and differences between the proposed algorithm and Hadoop resource scheduling algorithm, this paper makes a detailed study on three popular job scheduling algorithms: FIFO queue scheduling algorithm, Fair fair scheduling algorithm and Capacity algorithm based on computational performance. The advantages and disadvantages of each algorithm are analyzed in order to compare with the algorithm in detail. The number of node resources distributed in the cloud platform is very large, which inevitably leads to the emergence of unreliable node resources, and these nodes will have a great impact on the execution and scheduling of applications. In this paper, inspired by Bayesian cognitive model and guided by sociological trust relationship model, this paper first proposes a new cognitive trust model based on Bayesian method, and then applies this model to resource scheduling system. Theoretical analysis and simulation experiments show that the proposed method can effectively meet the trust requirements of cloud computing to node resources, and at the expense of less time cost, ensure the implementation of cloud computing tasks in a relatively secure node resource pool.
【学位授予单位】：山东师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.01
【参考文献】
相关期刊论文 前4条
1 王峰;;Hadoop集群作业的调度算法[J];程序员;2009年12期
2 查礼,徐志伟,林国璋,刘玉树;基于Simgrid的网格任务调度模拟[J];计算机工程与应用;2003年14期
3 张晓东;张培林;傅建平;王成;;基于Matlab/Fluent的协同仿真技术[J];计算机工程;2010年20期
4 孔邵颖;郭宏亮;;混合算法在网格任务调度中的应用研究[J];计算机仿真;2011年09期
本文编号：2261683
本文链接：https://www.wllwen.com/guanlilunwen/ydhl/2261683.html
