Machine Learning-Driven Container Scheduling for Edge-Empowered Microservices

This research investigates the integration of machine learning technology into microservices architectures for edge-enabled container scheduling strategies. Deploying microservices applications can be intricate, while container virtualisation technology solves environmental complexities. However, container scheduling strategies remain challenging, and machine learning techniques can be leveraged to rapidly and rationally adjust strategies for different services. A comparative analysis of container scheduling strategies with and without machine learning demonstrates a 60.6% improvement in scheduling speed. The evaluation includes two algorithms: K-Nearest Neighbours (KNN) and Bayesian Personalised Ranking (BPR). Results indicate that BPR outperforms KNN, achieving an 11.1% increase in speed, particularly for large-scale datasets. These improvements highlight BPR's effectiveness in edge computing environments, where containerised microservices operate closer to data sources and end-users. Additionally, the proposed control strategy minimises resource allocation errors and enhances system scalability. The findings provide a framework for integrating machine learning into edge-native microservices architectures, optimising container scheduling for edge computing scenarios.