Abstract:

The two-layer virtualized cloud architecture that deploys containers on virtual machines is increasingly used in cloud data centers. Aiming at the energy consumption problem of them, an algorithm named TUMS-RTC is proposed for workflow task scheduling. For the scheduling of a parallel workflow with deadline constraint, it divides the process into two phases：time utilization maximization scheduling and running time compression. TUMS reduces the numbers of virtual machines and servers needed to complete the workflow by making full use of the given time range. RTS shortens the working time of virtual machines and servers by compressing idle time of virtual machines, and finally the goal of reducing energy consumption can be achieved. The performance of TUMS-RTC is evaluated by using a large number of random workflows with controllable characteristics. Experimental results show that TUMS-RTC outperforms other algorithms with higher resource utilization rate, virtual machine number reduction rate and energy consumption saving rate. Moreover, it can handle large-scale workflows with high parallelism in cloud computing well.

Key words: cloud data center, energy consumption, workflow, task scheduling, virtual machine, container

摘要：

虚拟机上部署容器的双层虚拟化云架构在云数据中心中的使用越来越广泛。为了解决该架构下云数据中心的能耗问题，提出了一种工作流任务调度算法TUMS-RTC。针对有截止时间约束的并行工作流，算法将调度过程划分为时间利用率最大化调度和运行时间压缩两个阶段。时间利用率最大化调度通过充分使用给定的时间范围减少完成工作流所需的虚拟机和服务器数量；运行时间压缩阶段通过压缩虚拟机空闲时间以缩短虚拟机和服务器的工作时间，最终达到降低能耗的目标。使用大量特征可控的随机工作流对TUMS-RTC算法的性能进行了测试。实验结果表明，TUMS-RTC算法相较于对比算法有更高的资源利用率，虚拟机数量减少率和能耗节省率，并且可以很好地处理云计算中规模大且并行度高的工作流。

关键词: 云数据中心, 能耗, 工作流, 任务调度, 虚拟机, 容器