节点文献

关于提高网格计算性能和服务质量的几点研究

Research on Upgrading the Performance and Quality of Service in Grid Computing

【作者】 许智宏

【导师】 孙济洲;

【作者基本信息】 天津大学 , 计算机应用, 2004, 博士

【摘要】 网格是借鉴电力网的概念提出来的,它利用基于IP协议的宽带数字通信网络把分散在不同地理位置的电脑组织成一个“虚拟的超级计算机”,其中每一台参与计算的计算机就是一个“节点”,而整个计算环境是由成千上万个“节点”组成的“一张网格”,最终目的是希望用户在使用网格资源时就象现在使用电力一样方便。网格计算环境中资源分布共享、自治统一管理,以及资源的异构性和强动态性使得网格的性能和服务质量很难满足用户要求。本论文围绕提高网格环境的性能和服务质量开展了广泛的理论研究并进行了大量试验,取得了一些成果。利用蚂蚁算法具有的正反馈性和协同性、隐含的并行性及可扩展性适合于网络结构频繁改变的网格环境的特点,对基本蚂蚁算法进行了改进,设计了基于蚂蚁算法的网格任务调度策略。针对网格环境难于保证服务质量的问题,设计了一种实用的用户、资源和任务分类调度机制,在一定程度上实现了网格的QOS机制。为了仿真复杂的网格环境,我们在研究了现有的多种网格仿真环境的基础上设计了一种可扩展的网格仿真结构,适于研究不同的网格资源管理和任务调度方法。针对网格环境易用性差、应用开发困难的问题,提出了基于组件的应用开发环境,有利于提高网格的易用性,便于网格计算的普及应用,其中的RBAC用户管理机制有利于提高系统安全性,同时通过对组件的合理管理,有助于系统负载均衡和容错性的提高。

【Abstract】 Computational grid is based on power grid. It organizes distributed computers as a “virtual supper computer” by wide band digit network. Every computer is called a node, and all the nodes form a “Grid”. The target of grid is to make the users feel the use of grid is as convenient as using power grid. But the resources and user requirements in computational grid are more complex than power grid, and the distributed, autonomous, dynamic of resources make the performance and quality of service very hard to meet the user’s need.This paper made many theories study and do many experiments about upgrading the performance and service quality of grid environment, and get good results.We make use of the positive feedback, cooperativeness, parallelism and expansibility of ant algorithm which is very suitable to Grid, do some improvement on basic ant algorithm, and design the grid task scheduling method based on ant algorithm. For upgrading the service quality of grid, we use batch schedule and classify the users, tasks and resources in the grid, plan a practical QOS mechanism for Grid. For simulating the complex grid environment, we learn about the widely used grid simulate systems, design and realize an extendable grid simulate system, which is suitable to simulate different resource management and task scheduling methods. We find the application programming in grid environment is very difficult. So, we bring forward a grid application develop environment based on component technology. This environment is very easy to use, and by the proper management of components, the load balancing and fault tolerance of the system is easier.

  • 【网络出版投稿人】 天津大学
  • 【网络出版年期】2005年 03期
节点文献中: 

本文链接的文献网络图示:

本文的引文网络