Building a Cloud Infrastructure for Virtual Machine Scheduling in Datacenters
Jonathan Chua, Xunfei Jiang · 2024
While CloudSim and other simulation tools are widely used for algorithmic validation of workload scheduling algorithms in academic research, they often fall short in accurately replicating the complexities and dynamics of real-world cloud environments. In this paper, we propose a cloud infrastructure to provide a more substantial environment than traditional simulation-based approaches for testing and validating theoretical scheduling algorithms. The core of the platform is built using Ansible, an open-source tool known for its efficiency in automation and configuration management. Ansible is used to provision and configure a cloud infrastructure environment, laying the groundwork for a scalable and reproducible research test bed. Zabbix, a robust monitoring solution, is integrated across the infrastructure to gather detailed server and virtual machine metric data, providing critical insights for the heuristics used in various scheduling algorithms. The design and implementation of this platform is presented, while highlighting the practical challenges and solutions encountered. We demonstrate how Ansible, in conjunction with Zabbix, can be effectively used to create a cloud infrastructure capable of executing and evaluating scheduling algorithms, offering a closer approximation to real-world conditions compared to traditional simulations. The research platform developed in this paper addresses the limitation of simulators by leveraging bare-metal resources, offering a more realistic and tangible testing ground.