Adding a dynamic load balancing based on a static method in cloud via OpenStack

Abstract

With emerging Cloud computing technology over the Internet, accessing shared computing resources at low cost by consumers is increasing. Due to the massive demand for online services, load balancing is essential to optimize resource utilization and prevent overloading or underloading resulting from operating system updates, task operating time, server failure on the providers, and system failure due to hardware issues. To address these challenges of load balancing in cloud computing, various load balancing algorithms have been studied in static and dynamic modes. Although dynamic load balancing mechanisms have been proposed with their merits and demerits, none have been deployed based on the current static methods in the OpenStack platform, which is free of cost and has a giant community. Therefore, we contribute a dynamic solution based on the static method, Weighted Round Robin, in OpenStack. Our dynamic solution is able to balance the tasks considering CPU utilization criteria. As an evaluation, we analyzed the performance of the static method and our solution as a dynamic version of the standard method. Our experimental results showed less mean processing time and better performance in our solution than the static method. As a future work, we proposed three following methods. The first is considering multiple dimensions within our solution for dynamic load balancing in OpenStack to enhance efficiency. Next is using multiple clouds in different geographical regions to balance the workload among the providers in those regions. Last is a further design of our dynamic solution based on the static method, Weighted Round Robin in OpenStack using Machine Learning, with tracking current loads, CPU, Memory, and Network to calculate the suitable weights for each backend server to balance the users’ requests. Also, scaling out the server cluster’s resources can be employed through Machine Learning to provision a new virtual machine in a hectic time to serve the clients’ requests.

Date	1 Aug 2023
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Michel Kadoch (Supervisor)