【作者】 王洋；

【导师】 乔建忠；

【作者基本信息】 东北大学 ， 计算机系统结构， 2013， 硕士

【摘要】 网格系统的目标是实现大量异构资源的共享。由于网格资源具有动态性、异构性、分布性和自治性等特点,网格系统中资源分配已经成为了关键问题。经济学原理与传统的资源分配算法相比能更好的描述资源分配问题,所以很多研究者引入经济学的原理来解决网格环境下的资源分配问题。本文在已有研究成果的基础上,结合网格资源的特点,对ZI策略和固定标记策略进行了改进。提出一种基于供求关系和个人理性的定价机制(Supply and Demand based Individual Rational Pricing Mechanism SDIRPM),这个机制优化了基于连续双向拍卖的资源分配模型。本文的模型提高了基于经济理论的资源分配的效率。本文的主要工作如下：首先,通过对拍卖机制和个人理性算法的详细学习与研究,本文针对ZI策略和固定标记策略做出了改进。在两个策略中加入了历史值参考,使得ZI策略和固定标记策略更符合实际情况。其次,将Kaplan策略与GD策略结合形成一种新的个人理性算法,称之为基于供求关系和个人理性的定价机制。当交易平均价格较低时,供大于求,此时的Kaplan策略效率较高；当交易平均价格较高时,供小于求,此时的GD策略效率较高。再次,对基于连续双向拍卖的资源分配模型进行了优化设计。在网格环境中,通过引入连续双向拍卖原理,并结合基于供求关系和个人理性的定价机制,资源使用者可以获得计算资源完成任务,资源提供者可以提供空闲的资源来获得利润,这个过程可以提高资源的利用率。最后,基于GridSim建立了仿真实验,实现了多种分配算法和个人理性算法。将本文的基于供求关系和个人理性的定价机制与其他个人理性算法进行了对比,实验结果证明,本文的算法优于其他算法。更多还原

【Abstract】 The objective of grid system is to realize the share of numerous heterogeneous resources. Grid resources are dynamic, heterogeneous, distributed and autonomous, how to improve the efficiency of resource allocation for grid system has become a key issue. Resource allocation can be better described to utilize the principle of economics instead of traditional resource allocation algorithms. Therefore many researchers introduced economic theory to solve the problem of resource allocation in grid environment.In this paper, the improvement of the Zero Intelligence (ZI) strategy and the fixed mark-up strategy has been achieved based on the existed researches, and combined with the characteristics of grid resources. Thus a supply and demand based individual rational pricing mechanism(SDIRPM) is proposed, which optimized the resource allocation model based on the continuous double auction. The model in this paper improves the efficiency of resource allocation based on economic theory.The main work of this paper is as follows:First, on basis of detailed study and research on the auction mechanisms and individual rationality algorithms, ZI strategy and the fixed-markup strategy have been improved.To add the historical reference into two strategies, the ZI strategy and the fixed mark-up strategy are matched with the actual situation much better.Further, a new individual rationality algorithm which is called supply and demand based individual rational pricing mechanism is proposed, combined the Kaplan strategy with Gjerstad Dickhaunt (GD) strategy. The situation that average transaction price is low means that supply exceeds demand, Kaplan strategy has higher efficiency; otherwise, the situation that average transaction price is high means that demand exceeds supply, GD strategy has higher efficiency.Then, the optimization design of resource allocation model based on continuous double auction was implemented. The continuous double auction theory combined supply and demand based individual rational pricing mechanism was introduced to the grid environment. Resource users can obtain computing resources to complete tasks and resource providers can provide idle resources to obtain profit, this process improved the utilization rate of resources.Finally, the simulation experiments in the framework of GridSim were established to achieve various allocation algorithms and individual rationality algorithms. Supply and demand based individual rational pricing mechanism was compared with other algorithms in simulation experiments, and the results showed that algorithm in this paper is better than others.更多还原