基于QoS满意度增强的多资源协同预留策略

Enhanced QoS satisfaction based co-allocation reservation strategy in multiple resources

针对动态网格资源服务的不确定性问题,提出一种可量化分析资源服务QoS(quality of service)的多资源协同预留策略.该策略基于对运行在资源上的网格任务QoS指标分析,得出QoS满意度量化、归一化方法,建立资源服务QoS与预留容量之间的函数关系,并以市场经济环境为背景,分析任务费用约束下资源价格与预留容量之间的关系,求解得出可均衡负载的多资源节点协同预留方案.理论分析给出了策略的有效性证明和算法,仿真实验采用真实网格系统中的任务负载信息作为实验负载,在较大规模的模拟网格系统中检验了所提出的预留策略的性能表现.实验结果显示,该策略在接纳任务数、资源利用率和任务违约率方面的性能表现显著优于传统的预留策略.

Uncertainty for dynamic grid resource service, a quantitative analysis resources quality of service （QoS） of multiple resources co-allocation reservation strategy is presented in the paper. Based on the analysis of grid job QoS metrics running on the resources, the model obtain the QoS satisfaction quantification, normalization method and establish functional relationship between resource services QoS and reserved capacity. Additionally, the essay works out a balanced load multiple resources nodes co- allocation reservation strategy through the analysis of the relationship between the resource price and reserved capacity in the task cost constrains under the background of the market economy. The validity of the model and its algorithm are presented theoretically. The performance of the proposed strategy was simulated in a large-scale grid simulation system using the real task load of the practical grid system. The results show that the proposed co-reservation strategy outperforms traditional reservation strategy in terms of acceptance of task number, resource utilization and task delinquencies.