A Comparative analysis of openstack autoscaling engines: evaluating performance, scalability, and usability of heat and senlin under real-world workload patterns

Abstract

Cloud computing enables flexible and on-demand resource provisioning over the internet. Among its key features is auto-scaling, which is the automatic allocation and deallocation of resources based on the needs of the application without intervention by the user. Although auto-scaling is fairly mature in public cloud environments, its application in private clouds, particularly OpenStack meant for private cloud computing, is still limited. This gap in literature has been caused largely by a lack of performance data and comparative studies on OpenStack-based auto-scaling solutions. Manual scaling remains prevalent in most organizations but is often considered inefficient and difficult to manage. Public clouds pose challenges like higher costs, limited transparency, lower administrative control, and vendor lock-in. Hence, this paper presents a comparative analysis for performance evaluation of two native OpenStack auto-scaling engines Heat and Senlin, using ten realistic workloads generated using Apache JMeter. The ten workloads are evaluated in a fully-fledged OpenStack Zed environment, and results are analyzed using a Multi-Criteria Decision Analysis (MCDA) method. The conclusions provide practical guidance on selecting the best OpenStack auto-scaling engine based on workload patterns, thus filling a gap in the literature.