Here’s a little-known fact: did you know that BYU owns a supercomputer? It’s called the Mary Lou supercomputer, and it’s housed in the top floor of the Crabtree building. This computer is used by students and professors who require an enormous amount of computing power to do their research. It’s been used to compute complex physics simulations, analyze DNA strands, and even render 3D animated short films. Mary Lou is quite the busybody.
At 30 million dollars, Mary Lou’s computing prowess didn’t come cheap. But the scientific advancements and research made possible by this computer have been invaluable.
Wouldn’t it be great to have another one? What if I told you we could give Mary Lou a sister...for free?
Hear me out; It’s not as crazy as it sounds.
BYU campus has dozens of buildings, each containing potentially multiple computer labs that each house hordes of computers. All-in-all, there are thousands of desktop computers available. This army of computers provides for the computing needs of students throughout the day, but what do the computers do at night? Currently, nothing. They simply sit all night, running, but not doing anything valuable. Why not put those computers to work?
With the help of some open-source software and a bit of organizing we could harness the power of this otherwise unused army.
Here’s how it could work: A student or professor creates a research project that requires a lot of horsepower. Before sending their compute job to the supercomputer, their job would need to be split into small units of work, called packets, which can be worked on concurrently. Once the user submits their job, it goes to a central workstation which controls the flow of each of these packets through the buildings and computer labs on campus. The packets are distributed among the idle computers which perform the required computation, then send the results back to the owner. Even though each individual computer is not extraordinarily fast, as a whole, the entire virtual supercomputer is extremely fast because of its ability to process thousands of packets at once.
By merely installing a bit of software on each computer, we could have another supercomputer for next to nothing! We already own the hardware! So whaddaya say? What would you do with a free supercomputer?
A traditional supercomputer is not so different from the virtual one I proposed. It's normally a bank of thousands of processors, all networked together to form an extremely powerful machine. Complex simulations are still split into small chunks that each processor computes on its own. The results are compiled and analyzed for further research. The beauty of a virtual supercomputer is that the hardware is already owned by the university, which is undoubtedly the most expensive part of the whole proposal. Open-source software already exists that coordinates parallel processing in this way, so no additional cost would be required for purchasing software.
