混合云环境下基于失效感知的资源调度策略

Failure-Aware Resource Scheduling Policy for Hybrid Cloud

针对混合云的功能和复杂度逐渐增大而导致资源失效率增高的问题,提出一种提高混合云资源调度成功率的调度算法.首先,根据私有云资源失效规律特点,建立资源有效利用率模型和任务稳定性模型.然后综合考虑公共云和私有云的特点,建立基于失效感知的两层资源调度模型(2L-FARS),并使用建立的ST—LLF(任务稳定性阀值控制的最低松弛度优先调度算法)和DQPA(双队列资源提供算法)调度算法分别完成两层资源调度.最后使用failure traces和workload traces,对提出的策略进行验证.实验结果表明,该策略有效地减少了任务截止期违约率,并且在提高资源利用率的同时,一定程度上降低了任务执行总费用.

Aiming at increasing opportunities for resource failure caused by increasing function in hybrid cloud, the task stability model was constructed according to the features of failure. In order to evaluate resource utilization of private cloud, the effective utilization model of resources was built. In order to simultaneously ensure task stability and resource utilization of private cloud, the two layers resource scheduling mode was proposed based on failure aware. It included a hybrid cloud and private cloud scheduler, using the stability threshold--least laxity first task scheduling algorithm （ST-LLF） and dual queue provision algorithm （DQPA） scheduling algorithm respectively. Finally, using the real failure traces and workload traces, we evaluated the proposed resource provision policy to demonstrate their performance, resource utilization and violation rate. Experimental results show that the strategy effectively increases the resource utilization and reduces the violation rate.