面向实时网格任务的多策略资源协同分配模型

Multi-strategy resource co-allocation model for real-time task in computing grid

提出一种实时保证度的计算方法,用于评估特定协同分配方案对实时网格任务的截止时间保证度,并在此基础上设计和实现了一个多策略资源协同分配模型。该模型利用已有协同分配策略生成不同的分配方案,并选择截止时间保证度最优的方案为任务分配资源。仿真实验利用实际网格系统的任务负载,在多集群型计算网格测试床上检验了该模型的性能表现。实验结果显示,多策略协同分配模型能综合各种已有各类协同分配策略的特长,为实时网格任务的资源协同分配提供可靠性增强的截止时间保证。

A novel calculation method for the real-time guarantee level was suggested to evaluate the deadline guarantee level for the real-time task in the computing grid using a specific co-allocation scheme, and a multi-strategy resource co-allocation model was proposed and implemented based on this method. The model uses the existing co-allocation strategy to generate multiple co-allocation schemes and selects the scheme that provides the optimal deadline guarantee level to allocate the resource for the task. The performance of the proposed model was simulated in a multicluster computing grid test bed using the task load of the practical grid system. The results show that the proposed multi-strategy co-allocation model can integrate the strong points of the various existing co-allocation strategies, provide an enhanced deadline guarantee for the resource co-allocation of the real-time tasks in the computing grid.