混合云中基于截止时间和费用约束的调度方法研究*

针对混合云调度中私有云利用率不高和公有云费用偏高的问题,基于性能和费用目标提出了两个调度策略—截止时间优先和费用优先策略,建立了混合云中的任务和资源模型,能够根据用户提交的任务需求自适应选择合适的调度资源,对截止时间要求比较高的任务可以优先调度至公有云,对费用要求高的任务可以优先调度至私有云,而且两种策略均满足截止时间和一定的费用约束,因此相对于其它类似的基准调度方法,本文的两种调度策略在调度完成时间、费用、截止时间超出率和私有云利用率等方面均有很好的表现,尤其是当任务量比较大的时候,两种调度策略表现出更好的自适应性和优势。

Resource Scheduling Methods Based on Deadline and Cost Constraint in Hybrid Cloud

In hybrid clouds, private clouds are underutilized and public clouds are expensive. Aiming at mitigating these two problems, we propose two scheduling strategies (i.e., Time-First and Cost-First) based on performance and cost objectives. Time-First and Cost-First allow tasks to choose the appropriate scheduling resources to meet user demands. In particular, the task with precedence deadline can preferentially select public cloud, and the task demanding low cost can choose private cloud with priority. Both Time-First and Cost-First satisfy the constraint conditions of the deadline and cost. Experiments demonstrate that Time-First and Cost-First achieve good performance with respect to the completion time, the cost, the deadline time and the utilization rate of private cloud, compared with other similar scheduling methods. Especially when there are a lot of tasks, Time-First and Cost-First show better adaptability and have superiority relative to standard scheduling methods.