High Performance Computing (HPC) is large scale computing resources that are used to solve problems that require a significant amount of computing power. These often use very densely packed high performance computing nodes, connected by high speed networks. A common example is computer "clusters".
The following topics are described in more detail on this page.
The HPCRF is a university wide facility, managed by ICT, that provides a centralised location for dense computing configurations, such as computer clusters. The co-location of these computers provides benefits to both the users of the individual clusters, and the university, taking advantage of economies of scale.
An overview of the HPCRF was video recorded for the On Campus News' "Sneak Peek". The video can be viewed on the YouTube site.
For more information on HPCRF, please contact Jason Hlady.
Researchers considering purchasing new clusters should contact Mr. Hlady.
The U of S is a member of a consortium of western Canadian institutions providing high performance computing (HPC). The consortium is named WestGrid and participates in a national collaboration for high performance computing called Compute/Calcul Canada.
WestGrid and Compute Canada use a policy that permits all researchers to have access to high performance computing (HPC) resources across Canada based on the merits and needs of their projects. The strategy attempts to optimize the national investments in HPC by encouraging an appropriate spectrum of functionality, avoiding unnecessary duplication of facilities, and ensuring maximum usage of resources through sharing and global resource management.
The designated Principal Researcher from the U of S, and the U of S member on the WestGrid Executive Committee, is Dr. Raymond Spiteri from the Department of Computer Science. The contact for ICT involvement in WestGrid is Jason Hlady.
For queries regarding accessing HPC resources that are part of the current WestGrid consortium, contact WestGrid Support.
Note that the WestGrid Collaboration and Visualisation Facility is available for seminars and talks about WestGrid. It also provides some additional support for these and other researchers.
In May 2013, the university acquired a new Linux-based cluster to support research computing, to bridge the gap in needs between desktop or virtual machines and larger WestGrid/Compute Canada facilities. Jobs on Plato are managed by a "fair share" scheduling algorithm. For more information see Plato.
The University of Saskatchewan contributes a large scale disk storage facility (called Silo or Hopper) to WestGrid. Silo is used for file transfers, Hopper for file manipulations. Read more about the storage facility in this WestGrid article.
The High Performance Computing Centre is a Type A (college level) research centre of the College of Arts and Sciences. It exists to intensify current research and stimulate new research, encourage student supervisory collaborations between campus researchers, and provide a hub for students, instructors, researchers and employees at the U of S interested in high performance computing. Please contact the High Performance Computing Centre for more information.
ICT operates three facilities, Moneta, Socrates and Zeno, for training campus personnel on high performance computing. Instructors wanting to incorporate HPC into their classes can arrange access to the facilities for their students, but are responsible for any required software for their classwork.
These training facilities can be used to ensure that programs that will be submitted to WestGrid are capable of running in that environment. Debugging tools are available to test software before it submitted to WestGrid. However, these facilities are not intended for running production research software.
Moneta is to be used for those tasks that require a large amount of memory. It is equipped with 256 GB of RAM and 16 processors.
Socrates provides a cluster environment with 28 nodes with 8 GB of RAM each and 8 nodes with 32 GB of RAM each. Each node has 8 processors.
Zeno is an 8 node cluster connected by a high speed, low latency network called InfiBand. Each node has an NVIDIA Tesla GPU processor, for high speed computation on highly parallel jobs.
More information is available here.