About PittGrid
PittGrid is an ongoing project to create a virtual supercomputer by making use of unused computer resources across the Pitt campus.
When you leave your office or laboratory at the end of the day, if you aren't using your computer to run programs overnight, your CPU will be idle. Given the number of people working for the University and the high speed of most computers today, that's a lot of untapped resources! These resources could be better used to benefit the entire Pitt community.
PittGrid organizes these resources and makes them available for research, simulations, and computationally demanding projects from Pitt faculty and researchers. The aggregate power of several hundred machines is channeled to provide high computational power for people who have computationally intensive jobs.
Whether your research is environmental simulation, chemical modeling, disaster scenario simulation, or physics equation modeling, PittGrid can help you.
But PittGrid isn't just for researchers and high-throughput computing. You can run any processor-intensive programs written in C, C++, JAVA, Fortran, MATLAB or R. All of these types of jobs can also take advantage of the power of PittGrid.
PittGrid is a joint initiative of Office of the Provost and Computing Services and Systems Development (CSSD). And with the launch of Center for Simulation and Modeling (SAM), now PittGrid became part of SAM initiative.
How Does PittGrid Work?
To accomplish its task, PittGrid uses a Federated System Architecture, uniting disparate pieces of hardware (CPU's) and software databases from all over the University.
PittGrid makes use of Condor, a set of software tools that enable scientists and engineers to increase their computing throughput.
PittGrid also makes use of Globus, a middleware tier open source software toolkit used for building grids. Atop the API of this toolkit sits the Open Grid Services Architecture Data Access Integration service (OGSA-DAI for short), another middleware tier software which assists with access and integration of data from separate sources via the grid. This scheme allows access to many different types of databases.
Software aside, PittGrid is a series of hardware nodes, or CPUs, controlled by a single master server. The master server communicates with all of the nodes in the PittGrid, querying status and sending commands, divvying up the workload amongst the resources which are not already in use.
The master server is has two Dual Core Dual Xeon processors, running at 2.8GHz. The server has 8.0 GB of DRAM, and operates with a RAID 1 array of two hard disks (73.0 GB @ 15K RPM) for the operating system, and a RAID 5 array of four hard disks (330.0 GB @ 10k RPM) for data and file serving.
PittGrid's current CPU count is:
125 Linux CPUs 
300 Windows CPUs 
2 Mac/Solaris CPUs
Question? Comments?
Contact Senthil Natarajan:
senthil (AT) pitt (DOT) edu
Last updated: 09/11/14