Michael Christensen

BYU’s Fulton Supercomputing Lab (FSL) is a centralized campus resource whose mission is “to facilitate and enhance computationally-intensive research at BYU by providing reliable, state-of-the-art, high performance computing resources to faculty and students.” As part of that mission, Supercomputing provides large compute clusters and other systems for researchers to use as needed. If a researcher needs to do computationally intensive research that wouldn’t be feasible on a standard desktop, they can submit batch processing jobs to the lab’s systems and get their work done in a fraction of the time. A researcher’s code is then run in parallel on many processors on Supercomputing’s systems. For instance, Supercomputing has a particular user from Chemical Engineering who needs to do lots of runs of a large number of simulations. Each run would take him about two years on a single processor, but because he is able to use some of the lab’s resources, each of the runs takes about two to four days to complete. If more resources were available to him, he would be able to get his work done in a much shorter time and could get more runs through in the same amount of time.

Supercomputing wants to use the idle time of desktop computers on campus so that researchers can run jobs on those processors. There are many solutions out there already such as BOINC clients (e.g. Folding@Home and SETI@Home) that are great for certain research needs. The problem for the researchers is that these applications require certain programming models be followed and are often specific to a certain application. For instance, nobody has been able to find a BOINC-based client that does anything similar to what the aforementioned Chemical Engineering user is doing. Even if one existed, it would be very unlikely to benefit other users. Instead of using one of these custom solutions, the lab would like to provide the same operating system environment on the desktops that is already in use on its clusters. By providing the same environment, researchers would not be required to rewrite their code just to take advantage of the extra resources.

The project was a great success. Software has been written that runs on a computer and connects into a BYU Supercomputing cluster; thus expanding BYU Supercomputing resources. The software successfully runs on 32-bit and 64-bit Windows platforms (including Windows XP and Windows 7). It also works on both HP and Dell computers.

One original goal was to have the system deployed in both the School of Technology Information Technology (SOT IT) and Office of Information Technology (OIT) labs across BYU campus. The system has been successfully installed on a minimal number of OIT lab computers that OIT gave permission to test it on. OIT has been given a working version of the software to test, but are backlogged on work and haven’t had sufficient time to test it before deploying it across campus within the time-frame originally desired by the team. OIT will be receiving the final version of the software within the next few weeks. SOT IT has requested a special version of the software that will only run when the user logs off of the computer. That version will be complete and distributed to SOT IT within the next few weeks. Overall the project goals were reached. A system has been developed that successfully expands BYU Supercomputing resources.