|you:||Yes, yes, we know. The F1 guys use HPC to model their cars. Old news.|
|me:||True, but that isn’t my point.|
|me:||No. This is more about how HPC is like Formula 1.|
|you:||Because it’s *fast* right?
Alright Mr. Hoity-Toity Smarty-Pants, what’s your point?
|me:||Fair enough. I deserved that.|
Though comparatively unknown in America, the land of NASCAR, Formula 1 is arguably the most popular sport in the world other than football/soccer (I’m not getting in a debate here, and no, I don’t mean American football). Other than World Cup, it consistently beats out every other sporting event in terms of worldwide viewership, including the Summer Olympics. The speed and power captures the imagination of millions of people all over the globe every year as its 11 teams and 22 drivers traverse the world to race on the world’s most grueling tracks.
Formula 1 cars weigh about half that of a Cooper Mini. They race around tracks, some of which are downtown streets (e.g., Monaco) where the manhole covers have to welded down to keep the created up-force from lifting them free. But, Formula 1 has another exciting purpose. Many of the technological advances made for Formula 1 eventually make it into our everyday cars. Advances such as disk brakes, ground effects, anti-lock brakes, tire design, carbon fiber, rear-view mirrors, etc., courtesy of Formula 1 and the larger motorsport ecosystem. We all are safer because of the millions of dollars spent to create crazy vehicles capable of doing things like going from 0 to 160 kph (99.4 mph) and back to 0 in four seconds. These advances have even made it into other products.
So, let’s relate this to supercomputing.
There are really two different types of problems that supercomputing is trying to solve. First of all, there is the actual scientific or business problem for which the system was built. Every job submitted into an HPC cluster is really just a question for which one desires an answer. Secondarily (and perhaps more interesting), is the question of how one can accomplish the first in a faster, more efficient manner.
As I recently discussed, the high-end HPC systems are bespoke. They are designed and built specifically to run their workloads in the most efficient manner. Think of all the things that need to or have been considered:
- High-density machines
- Advanced cooling systems
- High-speed interconnects and networking
- GPUs, accelerators, and FPGAs
- Large-scale network storage
- Massive power requirements
- Intelligent workload placement
- . . .
And, the list just goes on and on. Essentially, these systems are on the bleeding edge of technology as they strive to squeeze out every last drop of computing power. In many cases, the question of how to make the system better is just as important as the work being done on it. So, necessity presents a problem to solve: how does one make it better?
Imagination is the creative force that through necessity yields solutions to resolve the issues that face us.
~ Steven Redhead
So, we struggle with this problem, and through effort we indeed make it better.
Then the magic happens.
The problems that are solved benefit not just us in the HPC community, but also the rest of the computer industry and the world. Like Formula 1, technology spillover occurs. What we learn from solving the HPC problem will improve personal computing, cloud computing, and even mobile devices.
All ideas grow out of other ideas.
~ Anish Kapoor
Now the only trick is to work out how best to share these ideas so they can grow. Before we can share outside of the HPC community, we need to share within it. How best can we, as the HPC community, better share ideas, knowledge, and insights for the benefit of all?
Image courtesy of Steve Gregory