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

Abstract

Cloud computing is a demanded field that opens the door for various possibilities due to its characteristics such as flexibility and on-demand availability. Basically, it presents an infrastructure for different kinds of services and tools via the public internet [1]. Auto-scaling, that is allocating and deleting resources without the involvement of the user, is one of the principal features in the domain of cloud computing. [2]. In domain of autoscaling on cloud systems, the demand for autoscaling systems, applications and services continues to surge, there is an increasing need for flexible and cost-efficient solutions to handle dynamic traffic patterns. Most of the time private cloud based autoscaling solutions are not used, because of the absence of proper information of the private cloud based autoscaling solutions and their performance. Specially in OpenStack who is the leading private cloud provider in the domain of cloud computing. As a result of this lack of knowledge and performance information of the private cloud based autoscaling solutions, the community are still using manual scaling methods when necessary or they are using public cloud based autoscaling solutions. However, there are several identified issues such as inefficiency and management challenges in manual scaling methods. When it comes to the public cloud based autoscaling solutions, there are set of challenges and cons for example, lack of cost-effectiveness, lack of transparency and administrative control, and vendor lock-in issues are few of them. As a solution for the above-mentioned issues, in this project we are giving the community an opportunity to get an idea about the OpenStack based autoscaling solutions (namely Heat and Senlin) and their behavior and performance against various kinds of practical workloads. As per our knowledge, there is not any previous literature that compares the different autoscaling solutions in OpenStack cloud. This work involves doing comprehensive research on the existing cloud-based auto-scaling solutions and the Zed version of fully functional OpenStack environment is implemented including the Heat and the Senlin projects. These autoscaling engines are tested with 10 different kinds of practical workloads which are generated using Apache JMeter and their performance metrics are recorded. These metrics are analyzed using MCDA method and the optimum autoscaling solution for each workload pattern is determined. We hope this work will contribute to the community who are interested in OpenStack based autoscaling to decide the proper autoscaling solution based on the nature of their workloads