Performance Analysis of Data Storage in a Hyperconverged Infrastructure Using Docker and GlusterFS

Abstract

Hyperconverged infrastructures use distributed file systems to store and replicate data between multiple servers while using the processing capabilities of those same servers to host virtual machines and containers. In this work, the distributed file system GlusterFS and the VMware ESXi hypervisor are used to build a hyperconverged system to host Docker containers in order to evaluate the performance of this system compared to a traditional approach in which data is stored directly on local disks of the servers. The performance of container persistent data storage is assessed using FIO (Flexible I/O Tester) storage benchmark tool with different workloads from Microsoft data centers and with multiple disk configurations in the hyperconverged system. The experimental results show that with a greater amount of writing operations and that handle large files, the performance in the hyperconverged architecture is better than in local storage, which presents an economically viable option in the implementation of storage for datacenters.