This is a Joint Initiative of The College of Engineering's
Faculty Academic Computing Committee and Electronic and Computer Service
Purpose
This system was purchased and deployed between May and July of 2006. Its primary purpose and function is to support those courses in the Engineering curricula that can use High Performance Computing to enhance the educational experience of undergraduates and graduates. Over the Summer 2010 semester, the College HPC cluster has been rebuilt as a Windows cluster optimized for the support of MATLAB jobs designed for parallel computing.
Custom Research Jobs
While the role of the College of Engineering HPC Cluster is to support instructional use research jobs will be allowed during periods of low instructional use. These jobs need to be designed and compiled to run on a Windows Server 2008 HPC cluster.
In order to submit jobs, the client machine must be running the HPC Job Manager, which can be obtained via SMB at \\engrhpc.ecs.psu.edu\hpc . The server is part of Engineering’s COEACCESS domain and will also allow logons from PSU_ENGINEERING accounts.
Please refer to Microsoft’s quick reference guide on how to submit jobs via the HPC Job Manager at http://technet.microsoft.com/en-us/library/dd379461(WS.10).aspx. A resource on the concepts of migrating Unix HPC applications to Windows can be found at http://www.pgroup.com/lit/whitepapers/pgi_whitepaper_unix2win.pdf.
Data files should be placed on \\engrhpc\data in a unique folder you create for your job and referenced according to their UNC path so it will be visible to the worker nodes. Please remember that this is a scratch data area. Once your job is finished please collect all pertinent data and clean up your residual raw data.
MATLAB
The Engineering cluster runs MATLAB for instructional use. MATLAB job submission to the ENGRHPC cluster requires that the client MATLAB computer be fully licensed and include a license for the Parallel Computing Toolkit. Please download the MATLAB configuration file (ENGRHPC.mat) for the Engineering HPC cluster at: "\\engrhpc\Matlab\Parallel Configuration" and import it into your MATLAB installation.
Additionally, you will need to have the Microsoft HPC Job Manager installed on your workstation. This can be obtained via SMB at \\engrhpc.ecs.psu.edu\hpc . The server is part of Engineering’s COEACCESS domain. All Engineering students as well as non-Engineering students taking Engineering courses have their Access accounts automatically added to the COEACCESS domain. The HPC will also allow logons from PSU_ENGINEERING accounts.
Large data sets and code can be posted to \\engrhpc\data so that it will be visible to all worker nodes. When submitting jobs you need to either have the ENGRHPC configuration as your default or reference it in your code.
HARDWARE CONFIGURATION
Head Node: Two Dual-Core 3.4GHz Processor System; 8 GB RAM, 1.2TB Local Disk Storage
Compute Nodes: 22 - Two Dual-Core 3.4 GHz Processors; 4 GB RAM, 40 GB Local Disk Storage
Gigabit Ethernet Connectivity Between Nodes
Please email with questions or recommendations.
Formerly Electronic and Computer Services (ECS)