May 22, 2013

Building Supercomputers with Raspberries

Nicole Hemsoth

At some point in the not-too-distant future, building powerful, miniature computing systems will be considered a hobby for high schoolers, just as robotics or even Lego-building are today.

That could be made possible through recent advancements made with the Raspberry Pi computers. These Linux-based computers are as small as a credit card and only cost about 25 to 35 dollars. Last year, the University of Southampton’s Simon Cox coupled 64 of these little systems together to form his own miniature supercomputer.

“As soon as we were able to source sufficient Raspberry Pi computers we wanted to see if it was possible to link them together into a supercomputer,” said Cox on how he created a small but powerful supercomputer on Raspberry Pi’s. “We installed and built all of the necessary software on the Pi starting from a standard Debian Wheezy system image and we have published a guide so you can build your own supercomputer.”

As seen in the diagram below, the diminutive processors are brought physically together by Legos while MPI serve as the interconnect system.

Joshua Kiepert, PhD candidate at Boise State, also got involved in building supercomputers out of the Raspberry Pi machines, completing a 32-node cluster that came in at a cost of just under $2000.

“In order to keep the cluster size to a minimum while maintaining ease of access,” Kiepert said in explaining how the computers were clustered, “the RPis (Raspberry Pis) were stacked in groups of eight using PCB-to-PCB standoffs with enough room in between them for a reasonable amount of air flow and component clearance.”

The advantage, according to Kiepert, to building his own system is the ability to outfit and customize based solely on his requirements. Per Kiepert, “by building my own cluster I could outfit it with anything I might need directly.”

Such customization included side-stepping the Pi’s micro-USB power port for a 5-volt pin that attached to the machines’ I/O ports. Further, Kiepert overclocked the Pis when he needed more processing power to run his simulations. With that said, the performance did still leave a little to be desired, according to Kiepert.

This story on fun, creative, and cost-effective ways to build supercomputers could actually represent a significant tool for today’s young learners who will grow up with computers, and in some cases supercomputers, all around them. Understanding the dynamics and logistics of high-performance computing early is important in maintaining interest in the field later in life.

“We want to see this low-cost system as a starting point to inspire and enable students to apply high-performance computing and data handling to tackle complex engineering and scientific challenges as part of our ongoing outreach activities,” Cox concluded.

Related Articles

Running Computational Fluid Dynamics in the Cloud

HPC and the True Cost of Cloud

CUDA-Designed Robot: It’s a Hit!

Share This