ML-Based Optimization Of Containerized Applications

Authors: Ananya Das

Abstract: The rapid adoption of cloud-native architectures has established containerization as the standard for deploying scalable applications. However, the inherent complexity of managing resource allocation, horizontal scaling, and placement in dynamic environments poses significant challenges to traditional heuristic-based management. Machine Learning (ML) has emerged as a transformative solution, offering the ability to predict workload patterns and optimize system parameters in real-time. This review explores the convergence of ML and container orchestration, focusing on how various algorithms—ranging from reinforcement learning to time-series forecasting—enhance performance while minimizing operational costs. By analyzing the current landscape of ML-driven auto-scaling, scheduling, and interference detection, this article highlights the transition from reactive to proactive resource management. The synthesis of these technologies not only improves Quality of Service (QoS) but also contributes to the sustainability of data centers by reducing energy waste. Ultimately, ML-based optimization represents a critical evolution in achieving truly autonomous, self-healing containerized ecosystems.

DOI: https://doi.org/10.5281/zenodo.19481640