Orchestration of Remote-Rendered Extended Reality Applications Leveraging Public Cloud Infrastructure

Abstract

Cloud rendering enhances extended reality (XR) by offloading heavy processing and graphics rendering to powerful remote servers. This approach improves device performance, graphical quality, and scalability while reducing the size and weight of standalone XR devices. Furthermore, it enables users with limited computing capacity devices to engage with traditionally heavy XR applications.

This thesis aims to explore the deployment and orchestration of remote- rendered applications using modern orchestration methods, with a focus on Windows-based containers requiring graphics processing unit (GPU) access. The primary aim is to compare the performance, scalability, and ease of deployment between traditional virtual machines (VMs) and containerized environments for remote-rendered Windows-native applications with graphical user interfaces (GUIs), thus challenging the traditional belief that a containerized application can only be headless.

The study involves deploying the developed prototypes in two ways, leveraging Amazon Web Services (AWS) global infrastructure: using simple Amazon Elastic Compute Cloud (EC2) VM instances and Kubernetes container orchestration in the cloud. Results indicate that Kubernetes offers performance comparable to traditional VM-based approaches while providing significant benefits in terms of resource utilization, deployment density, scheduling latency and scalability.

The research also discusses the challenges and limitations of using Windows operation system (OS) in a Kubernetes environment, emphasizing the difficulties of using WebRTC peer-to-peer communication protocol in the cloud. Future work includes exploring advanced networking solutions like AWS Wavelength and Kubernetes media gateways.

Overall, this thesis provides insights and practical guidelines for deploying latency-sensitive, Windows-based remote-rendered applications using Kubernetes, contributing to the field of DevOps and cloud computing.