基于模糊认知图的动态网格分布式资源调度

论文利用多代理的理论,通过网格节点自主选择任务来实现网格系统中的资源优化调度。由于各节点的自主性,对于任务分配方案将存在不同的支持度。论文采用模糊认知图,建立了对任务分配方案的支持度的协调控制策略,并在此基础上给出了一种智能的网格资源调度策略。考虑到各个网格节点的资源配置各不相同,提出了标准支持度的概念,保证了支持度协调策略的可行性和有效性。该调度策略无需存在处于上层的资源调度单元,各节点通过协调对任务分配方案的标准支持度即可实现网格资源的优化调度。该策略适用于分布式计算,并支持网格节点的动态变化,具有较好的实时性和鲁棒性。

A Distributed Resource Scheduling Based on Fuzzy Cognitive Map for Dynamic Grid Computation Systems

Based on the theory of multi-agent,the resource scheduling is implemented optimally by grid nodes choosing subtask independently in grid systems.Being autonomous,each grid node will give different support degree of task assignment.The fuzzy cognitive map is used to construct the support degree negotiation model,and based on it,an intelligent grid resource scheduling strategy is proposed in this paper.Considered that each grid node has different type of resource,the definition of standard support degree is proposed to insure the feasibility and validity of the support degree negotiation model.No upper level recourse scheduling unit is needed in the proposed scheduling strategy.Grid resource is allocated by grid nodes through negotiation on the basis of the interaction among the nodes due to the standard support degree of task assignment.This strategy is applicable for distributed computation,and supports dynamic change of grid nodes,which makes the grid computation system possess better real-time performance and robustness.