Exploring the Use of Remote GPU Virtualization in Low-Power Systems for Bioinformatics Applications

Carlos Reaño, Javier Prades, Federico Silla

Research output: Chapter in Book/Report/Conference proceedingConference contribution

2 Citations (Scopus)

Abstract

The large power requirements of current High-Performance Computing (HPC) deployments has motivated that several supercomputers and research proposals consider the use of low-power processor architectures instead of the traditional mainstream, but power hungry, x86 designs. However, many of these proposals do not consider the use of GPUs (Graphics Processing Units). GPUs are power efficient devices while in use and, therefore, considering them is aligned with the power-efficiency efforts that motivated the use of low-power processors. In this paper we explore the use of GPUs in several low-power processor architectures in the context of the bioinformatics domain. To that end, we make use of the rCUD remote GPU virtualization middleware to provide these low-power systems with access to GPU accelerators. In this study we leverage three different low-power processor architectures to execute three bioinformatics applications. The result of this exploration is that the Xeon D processor architecture is well suited for this purpose whereas other low-power architectures such as the Atom or ARM ones still need to be improved.
Original languageEnglish
Title of host publicationProceedings of the 47th International Conference on Parallel Processing Companion (ICPP '18 )
PublisherACM
Number of pages8
ISBN (Electronic)978-1-4503-6523-9
DOIs
Publication statusPublished - 2018

Fingerprint Dive into the research topics of 'Exploring the Use of Remote GPU Virtualization in Low-Power Systems for Bioinformatics Applications'. Together they form a unique fingerprint.

  • Cite this

    Reaño, C., Prades, J., & Silla, F. (2018). Exploring the Use of Remote GPU Virtualization in Low-Power Systems for Bioinformatics Applications. In Proceedings of the 47th International Conference on Parallel Processing Companion (ICPP '18 ) [8] ACM. https://doi.org/10.1145/3229710.3229733