A clustered virtual machine allocation strategy based on a sleep-mode with wake-up threshold in a cloud environment
- PDF / 865,799 Bytes
- 20 Pages / 439.37 x 666.142 pts Page_size
- 96 Downloads / 196 Views
A clustered virtual machine allocation strategy based on a sleep-mode with wake-up threshold in a cloud environment Shunfu Jin1
· Xiuchen Qie1 · Wenjuan Zhao1 · Wuyi Yue2 · Yutaka Takahashi3
© Springer Science+Business Media, LLC, part of Springer Nature 2019
Abstract The massive amount of energy consumed by cloud data centers is detrimentally impacting on the environments. As such, to work towards “greener” computing, in this paper, we propose a clustered virtual machine (VM) allocation strategy based on a sleep-mode with a wake-up threshold. The VMs in a cloud data center are clustered into two pools, namely, Pool I and Pool II. The VMs in Pool I remain awake at all times, while the VMs in Pool II go to sleep under a light workload. After a sleep timer expires, the corresponding VM will resume processing tasks only if the number of waiting tasks reaches the wake-up threshold. Otherwise, the sleeping VM will remain asleep as a new sleep timer starts. By establishing a queue with an N -policy and asynchronous vacations of partial servers, we capture the stochastic behavior of tasks with the proposed strategy, and derive the performance measures in terms of the average latency of tasks and the energy saving rate of the system. Furthermore, we provide numerical results to demonstrate the impact of the system parameters on the system performance. Finally, we construct a system cost function to trade off different performance measures, and develop an intelligent searching algorithm to jointly optimize the number of the VMs in Pool II, the wake-up threshold and the sleeping parameter. Keywords Cloud data center · Clustered VM allocation · Wake-up threshold · Sleep-mode · Average latency · Energy saving rate
1 Introduction The energy consumption in cloud data centers has been steadily increasing over the last few years (Zhou et al. 2018). As a result, the minimization of power and energy consumption in a cloud environment has become an urgent problem that is now receiving significant attention (Marek and Hoon 2018).
This work was supported in part by National Natural Science Foundation (Nos. 61872311, 61472342) and Hebei Province Natural Science Foundation (No. F2017203141), China, and was supported in part by MEXT and JSPS KAKENHI Grant (No. JP17H01825), Japan.
B
Shunfu Jin [email protected]
Extended author information available on the last page of the article
123
Annals of Operations Research
Applying a sleep mode is an effective method for reducing energy consumption in cloud data centers (Khojandi et al. 2018). For the purpose of minimizing energy consumption during runtime, Duan et al. (2015) presented a dynamic idle interval prediction scheme to estimate the future idle interval length of a CPU and thereby choose the most cost-effective sleep state. Chou et al. (2016) proposed a fine-grain power management scheme for data center workloads. This proposed scheme could dynamically postpone the processing of some tasks, create longer idle periods and promote the use of a deeper sleep mode. Luo et al. (2017) ba
Data Loading...
