时间约束下的云工作流动态混合资源调度

Dynamic Hybrid Resource Scheduling Based on Time-constrained Cloud Workflows

随着云计算的出现和云基础设施的快速部署,越来越多的大型工作流应用正在积极向云迁移.同时,如何在满足任务时间约束的前提下优化执行成本,提高资源利用率成为新的挑战.本文提出一种云工作流动态混合资源调度算法DHRS,不仅满足任务的时间约束而且在混合资源租用时取得较低的成本.首先,根据任务的优先级关系对任务进行预处理,基于概率升序对任务进行排序,并为子任务分配子截止日期;然后,依据顺序为工作流选择满足截止日期且成本较低的服务;最后,对每个服务动态选择预留资源和按需资源,基于预留资源的空闲时间段调度,进一步降低成本.在随机生成的不同的科学工作流上进行实验,并通过与现有算法对比,DHRS在满足时间约束并且降低执行成本方面具有一定的优势.

With the emergence of cloud computing and the rapid deployment of cloud infrastructure,more and more large-scale workflow applications are actively migrating to the cloud.At the same time,how to optimize the execution cost and improve the resource utilization under the premise of meeting the task time constraints has become a new challenge.This paper proposes a dynamic hybrid resource scheduling algorithm DHRS for cloud workflow,which can not only meet the time constraints of tasks but also achieve lower cost when renting hybrid resources.First,tasks are preprocessed according to the priority relationship of tasks,tasks are sorted based on ascending probability,and subtasks are assigned sub deadline;Then,according to the order,select the services that meet the deadline and have lower cost for the workflow;Finally,the reserved resources and on-demand resources are dynamically selected for each service,and the idle time period scheduling based on the reserved resources further reduces the cost.Experiment on different randomly generated scientific workflows,and compare with existing algorithms,DHRS has certain advantages in meeting time constraints and reducing execution costs.