Assessing the use of the BeeFS distributed file system on the OpenStack.

Abstract

OpenStack is one of the most used open source solutions to create private clouds for researching, testing and business purposes. As a cloud orchestrator, service aggregator, and IaaS provider, OpenStack controls pools of computing, storage and network resources in a data center. OpenStack services that provide some kind of storage (e.g., block, image, blob) rely on an underlying file system to provide the service and also for performance matters. This work presents the integration of the Beehive File System (BeeFS) as a distributed file system back-end storage for an OpenStack cloud, and a performance assessment considering Nova, a fundamental service in charge of managing virtual machines in the cloud. Experimental benchmarking results show that, in average, Nova performs 5 times faster using Ceph - the de facto distributed storage technology used in OpenStack - then when using BeeFS as its storage back-end. Although BeeFS reduces costs of creating distributed storage infrastructure, its integration with OpenStack is not complete which causes significant impact in the cloud applications performance.