2025 Articolo in conferenza

Dynamic and Forecast-Based Containers Autoscaling for Kubernetes with Reinforcement Learning

A. Lipari, G. Proietti Mattia, R. Beraldi

2025 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW)

Scarica PDF DOI: 10.1109/IPDPSW66978.2025.00169

Reinforcement learning Containers Predictive models Prediction algorithms Resource management

Abstract

Efficient resource management in Kubernetes is crucial for optimizing performance and cost in cloud computing environments. Traditional autoscaling methods react to workload changes but often fail to predict them, leading to underutilization of resources or performance degradation. This paper introduces a model-free Reinforcement Learning (RL) based autoscaler that leverages a deep Q-Network (DQN) alongside a Long Short-Term Memory (LSTM) network for workload forecasting. By predicting future request rates, our autoscaler proactively adjusts the number of bixvds replicas, ensuring compliance to Service Level Objectives (SLOs) while minimizing resource usage. Experimental results demonstrate that our approach outperforms standard Kubernetes auto-scaling strategies, achieving a resource consumption reduction of up to 10%% and improving performance levels up to 30%% compared to the default auto-scaling algorithm.

Citazione

Lipari, A., Proietti Mattia, G., & Beraldi, R. (2025). Dynamic and Forecast-Based Containers Autoscaling for Kubernetes with Reinforcement Learning. 2025 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW), 1081–1088. https://doi.org/10.1109/IPDPSW66978.2025.00169

BibTeX

@inproceedings{2025LipariDynamic,
  author = {Lipari, Alfredo and Proietti Mattia, Gabriele and Beraldi, Roberto},
  booktitle = {2025 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW)},
  title = {Dynamic and Forecast-Based Containers Autoscaling for Kubernetes with Reinforcement Learning},
  year = {2025},
  volume = {},
  number = {},
  pages = {1081-1088},
  keywords = {Reinforcement learning;Containers;Predictive models;Prediction algorithms;Resource management;Forecasting;Sustainable development;Surges;Long short term memory;Standards;kubernetes;resource management;machine learning},
  doi = {10.1109/IPDPSW66978.2025.00169}
}