Container scheduling optimization in a clustering environment

Abstract

The progress in container orchestration technology has led to a saturation level of microservices related solutions in recent years. Web applications typically exhibit a fluctuating workload that is challenging to estimate. Based on system parameters, the microservice design has the capability to dynamically alter the number of containers, either manually or automatically. There is a persistent issue regarding the adjustment of the number of containers in real time to match the current workload of the microservice. Several strategies have been utilized to tackle this difficulty effectively and swiftly, particularly in situations where there is a sudden and significant increase or decrease in the demand for services. Scheduling the container resources according to the demand is one of the most focused requirement of the industry. Service providers and researchers are attempting to decrease the expenses while upholding the Quality of Service. Among the autoscaling approaches rule-based approach, machine learning models, Fuzzy rules, control theory and hybrid approaches have been discussed in previous studies. To deal with the above research problem, a container scheduling strategy as a hybrid approach, based on machine learning techniques and rule based techniques will be discussed in this article. There is still potential for improvement in optimizing container scheduling based on fluctuating workloads while utilizing cluster resources. Based on this study and analysis the article has suggested a solution for Container Scheduling Optimization in Clustering Environment using a machine learning model with a rule-based approach as a hybrid approach to determine the scaling factor for the demand of the microservices. The suggested hybrid technique offers greater granularity in manipulating the scaling decisions. This is because the system predicts low-level parameters (demanding CPU Units) based on high-level data, rather than directly influencing the scaling factor. This research aims to assess the reliability of the proposed autoscaling strategy by analyzing the results obtained. This offers an implementation of autoscaling using microservices and a reference design for domains that are shared by the community.