论文部分内容阅读
传统数据网格调度算法容易陷入局部最优值和收敛速度过慢的问题。分析分层式数据网格的特点,对数据网格进行层次划分和节点角色二级划分。针对分层式网络调度模型,设计了一种基于节点博弈的分层式数据网格资源调度优化算法(CTDGRA算法)。该算法基于博弈论框架,将数据分布任务调度计划生成问题转变成静态数据任务与动态节点资源映射优化选取方案问题。兼顾数据任务间的依赖关系、节点域间的节点能力及节点的偏好行为,衡量各节点目标并获得全局最为有利或最为合理的方案的行为方案从而保证系统全局最优QOS。仿真实验表明,算法能激励普通节点贡献空闲能力的意愿,同时避免低性能节点成为资源获取的性能瓶颈,较好地提升系统的吞吐力。
Traditional data grid scheduling algorithm easily fall into the local optimal value and convergence speed is too slow. Analyze the characteristics of hierarchical data grid, and divide the data grid into two levels according to the hierarchy. According to the hierarchical network scheduling model, a layered data grid resource scheduling optimization algorithm (CTDGRA algorithm) is designed based on node game. Based on the game theory framework, the algorithm transforms the data distribution task scheduling plan generation problem into the optimization data selection solution problem of static data task and dynamic node resource mapping. Taking into account the dependencies among data tasks, node capabilities among node domains and node preference behavior, the goal of each node is to measure and obtain the best or most reasonable solution to the overall situation in order to ensure the optimal global system QOS. The simulation results show that the algorithm can stimulate the willingness of ordinary nodes to contribute idle capacity, and avoid low performance nodes becoming the performance bottleneck of resource acquisition and improve the system throughput.