Cloud Native Robotic Applications with GPU Sharing on Kubernetes

TL;DR

This paper explores deploying robotic applications on Kubernetes with GPU sharing, highlighting educational benefits, challenges faced, and proposing a cloud-native approach to improve scalability and resource management.

Contribution

It introduces a cloud-native deployment framework for robotic applications on Kubernetes, addressing GPU sharing and networking issues in educational settings.

Findings

Seamless simulation-to-real experience for students

Networking and GPU sharing challenges identified

Proposed alternatives for improved deployment in future courses

Abstract

In this paper we discuss our experience in teaching the Robotic Applications Programming course at ZHAW combining the use of a Kubernetes (k8s) cluster and real, heterogeneous, robotic hardware. We discuss the main advantages of our solutions in terms of seamless simulation-to-real experience for students and the main shortcomings we encountered with networking and sharing GPUs to support deep learning workloads. We describe the current and foreseen alternatives to avoid these drawbacks in future course editions and propose a more cloud-native approach to deploying multiple robotics applications on a k8s cluster.