数据网格副本管理策略的研究

发布时间：2018-05-28 03:13

本文选题：网格 + 数据网格　；参考：《燕山大学》2014年硕士论文

【摘要】：网格是一种新兴的技术，它将众多的信息和资源的孤岛连接了起来，人们通过网格使工作、生活等更加的便利。网格中的计算网格是一个计算密集型应用平台，而数据网格是由计算网格衍生而来，它是一个分布式异构的数据密集型和计算密集型的平台，但又与计算网格没有严格的分界线。数据网格是一个主要以数据为主要资源的网格平台，数据网格中的核心技术即副本管理策略，有效的副本管理策略可以减少网络的传输时延、网络带宽的消耗等。首先对当前的副本管理策略现状进行了分析，针对当前数据副本管理策略的效率不高、性能不佳等特征，对副本放置、副本替换以及副本创建中的副本放置三方面进行了分析和研究。合理地利用网格的环境并作出有效的副本创建可以提高数据网格系统的性能，通过副本的定位和选择可以找到最佳的副本，副本替换策略是副本管理技术的核心，因为每个节点的存储单元的空间都是有限的，一个好的副本替换策略可以提高作业的执行速度、存储单元的利用率以及减少带宽的消耗。而副本的一致性则有效的解决了两个及以上的副本一致性问题。其次，对DHRA副本管理策略的副本选择算法、副本创建算法进行了研究，总结了它们的缺点，考虑了其它的因素提出了改进的副本选择算法以及改进的副本创建算法。再次，对DHRA副本管理策略的副本替换算法进行了研究，论证了在副本删除时只考虑副本历史访问时间的不足。提出了基于副本权重和副本代价的改进的副本替换算法。将这些算法进行了整合提出了新的副本管理策略，并用java语言实现了该策略。最后，，采用了Optorsim数据网格模拟工具对LWLC数据副本管理算法进行了仿真模拟实验，也对DHRA副本管理策略、LRU副本管理策略、LFU副本管理策略进行了仿真模拟实验。并对实验结果进行了对比分析，分析结果验证了LWLC数据副本管理策略的有效性。
[Abstract]:Grid is a new technology. It connects many isolated islands of information and resources. People use grid to make work and life more convenient. The computational grid in the grid is a computational intensive application platform, and the data grid is derived from the computing grid. It is a distributed heterogeneous data-intensive and computation-intensive platform. However, there is no strict boundary with the computational grid. Data grid is a grid platform with data as the main resource. The core technology of data grid is replica management strategy. Effective replica management strategy can reduce the transmission delay and bandwidth consumption of the network. Firstly, the current status of replica management strategy is analyzed, aiming at the low efficiency and poor performance of the current data replica management strategy, the replica is placed. Replica replacement and replica placement in replica creation are analyzed and studied. Making reasonable use of grid environment and making effective replica creation can improve the performance of data grid system, and the best replica can be found by locating and selecting replicas. The replica replacement strategy is the core of replica management technology. Because the storage cell space of each node is limited, a good replica replacement strategy can improve the job execution speed, memory cell utilization and reduce bandwidth consumption. The consistency of replica effectively solves the problem of two or more replica consistency. Secondly, this paper studies the replica selection algorithm and replica creation algorithm of DHRA replica management strategy, summarizes their shortcomings, and proposes an improved replica selection algorithm and an improved replica creation algorithm considering other factors. Thirdly, the replica replacement algorithm of DHRA replica management strategy is studied, and the deficiency of replica history access time is demonstrated when the replica is deleted. An improved replica replacement algorithm based on replica weight and replica cost is proposed. A new replica management strategy is proposed by integrating these algorithms and implemented in java language. Finally, the Optorsim data grid simulation tool is used to simulate the LWLC replica management algorithm, and the DHRA replica management strategy is also simulated. The experimental results show that the LWLC replica management strategy is effective.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.02

【参考文献】