A Cloud Resource Prediction and Migration Method for Container Scheduling

Abstract

With the continuous evolution of cloud-native, more and more applications are deployed on the container. As the business platform runs many container instances and has complex dependency relationships, the load of cloud resources fluctuates due to service status, resulting in task scheduling difficulties and affecting service stability. In this paper, we propose a container resource migration scheduling algorithm called UVPOC, and establish a two-level scheduler mechanism to monitor global real-time resources, choosing LTSM to predict the trend of the dominant resource utilization and determine whether to perform container migration or Virtual Machines (VMs) pre-boot to implement resource allocation. In addition, we choose CloudSim open source tool for simulation experiments. The results show that our UVPOC algorithm can improve the global resource utilization of containers and virtual machines, and reduce the energy consumption of data center resources.