AbstractsComputer Science

GPU cluster for acceleration of scientific and engineering applications in the context of higher education

by Matthew Newall

Institution: University of Huddersfield
Year: 2015
Keywords: LB2300 Higher Education; QA75 Electronic computers. Computer science; TA Engineering (General). Civil engineering (General)
Record ID: 1390671
Full text PDF: http://eprints.hud.ac.uk/23746/


Many fields of research now rely on High Performance Computing (HPC) systems which can process ever larger datasets, with increasing accuracy and speed. Many universities now provide a HPC service. Following the trend over the past few years of the worlds fastest supercomputers being accelerated using Graphical Processing Units (GPUs), there is a growing interest in the use of GPUs in Higher Education Institutions. The characteristics of GPUs make them excellently suited to any task exhibiting a high level of data parallelism. Recent developments in GPU technologies have focused on improving performance and integration in HPC, and for processing data other than display graphics. To investigate the benefits such a system could have to the University of Huddersfield, a small GPU cluster has been deployed. The intention behind this thesis is to detail the deployment of the system and to demonstrate, through case studies, the required effort a potential user could expect in order to take advantage of it. As a result of this work it can be demonstrated that even a modest GPU cluster can be of benefit to the University. The cluster is helping our researchers to analyse complex data using visualisation, and accelerating data processing.