混合计算环境中截止期约束下的科学工作流调度策略

Scheduling Strategy for Deadline-Sensitive Scientific Workflow on Hybrid Grid-Cloud Infrastructure

整合云和网格基础设施,增强科研机构现有网格系统的计算能力并向应用提供截止时间保障的服务是科学研究领域的热点。在这种"网格-云"混合计算环境中,对何时租借云虚拟资源以及如何租借做出有效决策是一个难题。现有的一些调度策略主要在网格资源静态能力特征的基础上,以作业等待时间作为决策依据,缺乏对资源动态服务能力的有效评估,无法保证科学应用的截止时间需求。本文提出了一种混合环境下的科学工作流执行系统架构并对其核心组件进行了阐述。针对其中的工作流调度问题,利用随机服务模型建模已有网格系统中的资源的动态服务能力,以任务违约风险作为是否租借外部虚拟资源的判断指标,提出了一个科学工作流调度算法HCA_SASWD。实验结果表明,HCA_SASWD相对于其他算法,能有效保证用户的截止时间要求,为需要提供截止时间保障的系统架构提供了参考。

In a hybrid grid-cloud environment, the user has elasticity provided by public cloud resources that can be aggregated to the existing grid resource pool as necessary. However, the integration opens new problem,such as：which are the best resources to loan from a public cloud based on current QoS requirement？ Existing scheduling strategies based on static resource information cannot absolutely guarantee the successful completion of scientific workflow and satisfy the user＇deadline requirements. In this paper,we propose an infrastructure, which is able to manage the execution of scientific workflow. Aimed at the scheduling of deadline-sensitive scientific workflow, the stochastic service model is adopt to model dynamic service capacity of grid resource and a metric called Default Risk of Task （DST） is offered to judge whether virtual resources should be loaned from cloud providers or not. Thus a scheduling algorithm （HCA_SASWD） is put forward. The experimental results show that HCA_SASWD achieves better performance than other algorithms on user＇s deadline guarantee.