Reinforcement Learning Based Vertical Scaling for Hybrid Deployment in Cloud Computing

Cao，Jianqi1; Li，Guiying1,2; Yang，Peng1,3

10.1007/978-981-99-1549-1_32

2023

1865-0929

1865-0937

Communications in Computer and Information Science

1801 CCIS

408-418

To maximize the CPU utilization of the server, offline tasks are usually deployed to the same server where the online service is running. Considering the necessity to ensure the service quality of online services, it is common practice to isolate the resources of online services. How to set the resource quota for online services not only affects the service quality of online services, but also affects the number and the stability of offline tasks that can be run on the server. Traditional rule-based methods or prediction-based methods will cause over-provision and fail to consider the stability of offline tasks, which often cannot achieve stability and efficiency. In this paper, reinforcement learning is proposed for the first time to solve the hybrid deployment of online services and offline tasks and dynamically adjust the resource quota of online services more effectively. Compared with the original state of the server, our proposed method reduces CPU idleness rate by 35.32% and increases CPU resource utilization rate by 3.84%.

Cloud Computing Reinforcement Learning Vertical Scaling

First ; Corresponding

English

2-s2.0-85161429146

Scopus

1.Guangdong Provincial Key Laboratory of Brain-Inspired Intelligent Computation,Department of Computer Science and Engineering,Southern University of Science and Technology,Shenzhen,518055,China
2.Research Institute of Trustworthy Autonomous Systems,Southern University of Science and Technology,Shenzhen,518055,China
3.Department of Statistics and Data Science,Southern University of Science and Technology,Shenzhen,518055,China

Cao，Jianqi,Li，Guiying,Yang，Peng. Reinforcement Learning Based Vertical Scaling for Hybrid Deployment in Cloud Computing[C],2023:408-418.