Research Article
Towards a GPU-Accelerated Open Source VDI for OpenStack
@INPROCEEDINGS{10.1007/978-3-030-99191-3_12, author={Manuel Bentele and Dirk von Suchodoletz and Manuel Messner and Simon Rettberg}, title={Towards a GPU-Accelerated Open Source VDI for OpenStack}, proceedings={Cloud Computing. 11th EAI International Conference, CloudComp 2021, Virtual Event, December 9--10, 2021, Proceedings}, proceedings_a={CLOUDCOMP}, year={2022}, month={3}, keywords={VDI OpenStack Desktop virtualization Remote access}, doi={10.1007/978-3-030-99191-3_12} }- Manuel Bentele
Dirk von Suchodoletz
Manuel Messner
Simon Rettberg
Year: 2022
Towards a GPU-Accelerated Open Source VDI for OpenStack
CLOUDCOMP
Springer
DOI: 10.1007/978-3-030-99191-3_12
Abstract
Starting from summarizing preexisting work and technologies, this paper introduces the necessary considerations and steps to develop a fully Open Source Virtual Desktop Infrastructure (OSVDI) on top of OpenStack. Unlike other already available open-source solutions, a Virtual Desktop Infrastructure (VDI) with access for 3D or video rendering using Graphics Processing Units (GPUs) caters to various use cases like GPU-accelerated desktop environments, remote visualization, or large-scale desktop virtualization. Such use cases require the integration of special-purpose hardware in cloud servers and the sharing of devices as well as resource scheduling, a user interface for session or resource selection, and efficient remote transport. The envisioned OSVDI is yet in its infancy growing from previous work to provide efficient large-scale remote access to existing PC pools in the bwLehrpool service for various teaching and learning environments. This preliminary cloud provides insights and experience to design the necessary (additional) OpenStack components and configurations.
