English

Taming the Memory Beast: Strategies for Reliable ML Training on Kubernetes

Distributed, Parallel, and Cluster Computing 2024-12-30 v2 Machine Learning

Abstract

Kubernetes offers a powerful orchestration platform for machine learning training, but memory management can be challenging due to specialized needs and resource constraints. This paper outlines how Kubernetes handles memory requests, limits, Quality of Service classes, and eviction policies for ML workloads, with special focus on GPU memory and ephemeral storage. Common pitfalls such as overcommitment, memory leaks, and ephemeral volume exhaustion are examined. We then provide best practices for stable, scalable memory utilization to help ML practitioners prevent out-of-memory events and ensure high-performance ML training pipelines.

Keywords

Cite

@article{arxiv.2412.14701,
  title  = {Taming the Memory Beast: Strategies for Reliable ML Training on Kubernetes},
  author = {Jaideep Ray},
  journal= {arXiv preprint arXiv:2412.14701},
  year   = {2024}
}

Comments

4 pages