华南理工大学学报(自然科学版) ›› 2009, Vol. 37 ›› Issue (4): 13-17.

• 计算机科学与技术 • 上一篇    下一篇

基于三方博弈的网格资源协同分配模型

肖鹏 胡志刚   

  1. 中南大学 信息科学与工程学院, 湖南 长沙 410083
  • 收稿日期:2008-04-22 修回日期:2008-05-20 出版日期:2009-04-25 发布日期:2009-04-25
  • 通信作者: 肖鹏(1979-),男,博士生,主要从事网格计算研究. E-mail:xpeng4623@yahoo.com.cn
  • 作者简介:肖鹏(1979-),男,博士生,主要从事网格计算研究.
  • 基金资助:

    国家自然科学基金资助项目(60673165)

Co-Allocation Model of Grid Resource Based on Three-Side Game

Xiao Peng  Hu Zhi-gang   

  1. School of Information Science and Engineering, Central South University, Changsha 410083, Hunan, China
  • Received:2008-04-22 Revised:2008-05-20 Online:2009-04-25 Published:2009-04-25
  • Contact: 肖鹏(1979-),男,博士生,主要从事网格计算研究. E-mail:xpeng4623@yahoo.com.cn
  • About author:肖鹏(1979-),男,博士生,主要从事网格计算研究.
  • Supported by:

    国家自然科学基金资助项目(60673165)

摘要: 针对费用约束下的网格资源分配问题,提出一种基于三方博弈的资源分配模型.该模型通过引入一类“虚拟资源中介”实体,将资源分配问题转化为由用户任务、虚拟资源中介、资源提供者三方组成的博弈过程.在合作博弈模型中,系统通过批发价格来保证资源提供者的收益;在非合作博弈模型中,系统通过零售价格杠杆来优化资源利用率和用户费用.文中以效用函数为基础给出了模型的有效性证明和求解算法,并通过实际网格任务负载和典型网格测试床模型检验了该三方模型的性能.仿真实验结果显示,对于具有费用约束的网格任务,相比于传统的资源分配模型而言,基于三方博弈的协同分配模型能增加约16%的资源收益和24%的资源利用率,从而有效地提高了任务的执行效率.

关键词: 计算网格, 协同分配, 效用函数, 计算经济, 博弈论

Abstract:

To deal with the resource co-allocation with the constraint of user budget in grid environments, this paper proposes a resource co-allocation model based on the three-side game. In this model, a new type of entity named Virtual Resource Retailer is introduced to transfer the typical resource co-allocation process into a three-side game process consisting of the user task, the virtual resource agency and the resource provider. In the cooperative game model, the benefit of resource providers is guaranteed by setting a wholesale price, while in the non-coopera- tive game model, the resource utilization and the user's cost are optimized by using a retail price lever. The validity and solution of the three-side model are then presented based on the utility function, and the effectiveness and performance of the model are verified by using the real grid workload based on a typical grid test-bed model. Simulated results show that, when dealing with the grid tasks with a cost constraint, the proposed three-side co-allocation model outperforms the conventional economy-based one by about 16% in term of resource benefit, and by 24% in term of resource utilization rate. Thus, the execution efficiency of grid tasks is effectively improved.

Key words: computational grid, co-allocation, utility function, computational economy, game theory