基于博弈论健壮性增强的资源分配模型

Game-theory-based robust-enhanced model for resource allocation in computational grids

如何保障用户QoS是动态、不确定网格环境中资源分配面临的一个新问题.针对网格资源的不可靠性特点,提出了健壮性增强的作业执行服务的概念及其确定方法,基于该概念提出了健壮性增强的离线资源分配模型,通过将作业执行服务中对应的多个网格节点资源分配给作业冗余执行以增强其健壮性.基于该模型将资源分配描述为一个非合作博弈过程,求得了帕累托最优分配方案,并提出了一种健壮性增强的离线资源分配策略.实验结果表明:该分配模型兼顾了效率和公平,增强了作业执行的健壮性,并较好地评估资源的信任行为.

Howto guarantee user＇s QoS for resource allocation in dynamic and uncertain Grid environment is a new challenge. Considering unreliability of grid resource, a definition called Robust-enhanced Job Execution Service （RJES） and its constructing method is proposed. A Robust-enhanced Model for Resource Off-line Allocation （RMROA） is derived from RJES. Robustness of user＇s job is enhanced through allocating some resource in each Grid site in RJES to the jobs redundantly. Based on RMROA, resource allocation is modeled as a non-cooperative game. A Pareto optimal allocation scheme is obtained and Robust-enhanced Off-line Allocation Policy （ROAP） is presented. We conduct extensive experiments, and the results show that RMROA not only looks after both efficiency and impartialness, but also enhances robustness of job execution and provides good mechanism for evaluating trust.