【作者】 张涛；

【导师】 须文波；

【作者基本信息】 江南大学 ， 计算机应用技术， 2006， 硕士

【摘要】 网格计算已成为当前研究的一个热点。在网格系统中,资源调度是一个难题。优秀的资源调度策略,可以减少任务的执行时间和费用、提高服务质量、增加整个网格系统的吞吐量。本文就网格的计算经济模型中计算资源调度算法进行探讨。全文主要围绕资源搜索算法、服务质量、负载平衡和吞吐量几方面展开。为方便资源提供者和资源消费者表达自己的需求,提出资源与任务描述语言;为实现在大量的计算资源中高效率地搜索到优秀资源,设计了资源搜索算法。该算法结合了Cache系统原理、局部性原理、蚂蚁算法和遗传算法的优点,大大地改善遗传算法的性能,能够动态高效地搜索计算资源。在确保服务质量方面,根据网格的计算经济模型特点,建立资源提供者信誉度模型和资源消费者信誉度模型。该模型既能确保交易对客观因素的要求,又能实现对交易者主观态度的评估,能够确保服务质量。通过将任务分类和提供资源参考价格的策略,在不影响其它指标的情况下缓解负载压力,提高系统资源利用率。在网格的计算经济模型中计算资源是记费的,本文优秀资源的一个主要评价指标就是资源价格,这可以确保搜索到价格合理的资源,可以最大限度地降低费用。对于吞吐量,将大资源分配给大任务可以减少调度次数,节省调度时间,从而提高吞吐量。另外,资源提供者和资源消费者的信誉度也与吞吐量密切相关,本文就信誉度与吞吐量的关系进行了探讨,并证明了将信誉度较大的资源提供者的资源分配给信誉度较大的资源消费者,可以提高系统吞吐量。最后,运用网格模拟器GridSim对算法进行测试,实验结果证明本文算法是有效的。更多还原

【Abstract】 Grid computing has become the hot target of research. In Computational Economy Model of Grid, resource scheduling problem is a hard problem. The bad resource scheduling will increase the time and cost needed for tasks, affect the Quality of Service(QoS), lower the throughput of grid.This research paper is talking about the resource scheduling algorithm based on Computational Economy Model of Grid. The soul of this paper is about efficient resource searching algorithm, QoS, load balance and throughput. In order to provide convenience for providers of resource and consumers of resource to express their demands easily, bringing forward describing language of resource&task. In order to search resource efficiently, a resource searching algorithm is designed. This algorithm combines the advantages of Cache system, Ant Colony Algorithm, locality principle and Genetic Algorithms, can search resource efficiently. Come to the strategy of QoS, Model of Resource Provider Credit and Model of Resource Consumer Credit are set up, according to characteristics of Computational Economy Model. These models not only can ensure the demands of deal, but also can evaluate attitude of dealers, can ensure QoS. This paper takes advantage of providing reference price and task sort to put off pressure of load and improve the resource utilization rate, not affecting other factors. In the Computational Economy Model of Grid, the resource is not free, the resource price is considered into searching algorithm, so the price of resource is reasonable. It is found that allocate large resource to large task can increase throughput because of reducing scheduling times. On the other hand, Resource Provider Credit and Resource Consumer Credit affect throughput greatly, this problem are discussed and an answer are given. At the end, use GridSim to simulate the algorithm and the result shows that the algorithm is valid.更多还原