Why You Should Give A Damn About Supercomputers

When you hear the word “supercomputer,” what are the first mental images that pop into your head?

If you’re like most of our readers who work around these systems, you probably conjured a vision of long lines of what look like refrigerators, all neatly cabled together. But chances are, if you’ve never navigated the noisy aisles of a large-scale datacenter, the image your mind’s eye crafted was likely a cross between a 1950s robot and a shiny cube with blinking lights.

Not that you needed to be told, but if your view was the latter, that is science fiction. And outdated science fiction at that. There was a long stretch in film where supercomputers were depicted visually–something that has, interestingly, slowed significantly in recent decades for whatever reason. But that’s another conversation, even if it explains the cinematic quality to show the non-technical an imagine of supercomputers. The point is, the contemporary reality of supercomputing is much more sleek and nondescript (underneath the painted cabinets) externally, but the symphony of activity happening inside is anything but ho-hum.

And for the record, there is nothing happening inside those machines that threatens the human race. Just making that clear from the outset.

For those who exist outside the high performance computing bubble (to whom none of what is here will come as a surprise) or, for that matter, for anyone who is generally disinterested in technology trends as a whole, supercomputers are curiosities–mythical metal beasts that answer esoteric questions, whip humans at Jeopardy! or chess, and at their worst, grow listless in their all-knowing sentient states and suddenly decide to take over the world. Their relevance is lost in popular culture translation.

In fact, the word “supercomputer” itself comes with so many connotations that have little to do with its mainstream, practical purpose that in conversations some tend to opt for the much more balanced “high performance computing.” It sounds much more grounded that way, and doesn’t come with all the science fiction.

The fictional stories about supercomputers (Skynet, Mother, HAL, and so forth) rarely depict benevolent systems, and in fact, the more aware they are, the more dangerous they become. This is not to say there isn’t something to be concerned as we create ever-smarter machines, but that’s the not the point just now. And to be fair, this same sort of “bad press” around a technology is happening with artificial intelligence. To make it worse, if one says, “yes, well, to do artificial intelligence at proper scale, we will need powerful supercomputers” to anyone outside the bubble, eyes widen and heads shake. Add in the idea that (eek!) robots will be involved in that mix and there will be utter pandemonium.

This is because there has been little success in making the general public understand what supercomputing and artificial intelligence are—what these things do. It’s far easier and much better clickbait to talk about how supercomputers were used for massive Bitcoin mining or how artificial intelligence leaps are taking away jobs.

It’s all just missing the point. Same goes for the “scary” applications running on monolithic supercomputers, including (gasp!) artificial intelligence jobs.

So where is the magical balance to be struck between sensationalizing that which is justifiably sensational (for crying out loud, there is a machine in Tennessee that is capable of doing 20,000 trillion calculations per second) and fear-mongering at the expense of undervaluing and glossing over the real meaning of these technologies?

20,000. Trillion. Math Operations. And within the space of a single second.

Can you even take that in?

Can our flimsy, relatively high-latency, low-bandwidth bundles of neurons pull together enough context to process that number into something concrete with a practical value assigned?

Is it possible for anyone to really look at raw performance figures and make connections about what mathematical floating point capability has to do with the real world?

It’s not difficult to understand that the applications running on these machines are valuable. For instance, we may know that we are able to simulate entire planetary interactions, climates from ancient times, or earthquakes in ultra-high resolution—but can the wider world and general public (who are paying for many of these, by the way—that’s the important bit) grasp why the systems themselves (okay, that implies they are a “they”) are worth paying attention to? What they mean for competitiveness, for progress, for possibility?

In fact, is that the problem? Is the source of the disconnect between these trillion floating point figures and our value or emphasis on supercomputing simply that there is no simple way to communicate

Is the missing story in the public supercomputing narrative simply that we have no frame of mental reference for what this capability means?

At all of these points, the story falls apart.

And that’s a shame, because not only are there practical benefits from what supercomputers and their complex applications lend to the world, it’s getting harder than ever to justify the costs for these massive machines.

For some context, consider that the federal government in U.S. set aside $425 million for three national labs—providing a total of three very large systems (although there are smaller machines worked into those deals to prepare users and programmers for the big daddies once they arrive). One of the top systems in the United States, the Titan supercomputer at Oak Ridge National Lab, was an addition to the existing system with the upgrade alone totaling $60 million.

In the U.S and many other countries, large-scale computing resources are government funded. The Department of Energy in the United States leads the way, but many other scientific agencies and departments around the world push their supercomputing prowess with public/private partnerships or with universities and research centers footing the bill. With those costs in mind, remember, too, that big companies are also buying supercomputers. Just as with the political argument, the point behind the spend is to lend competitiveness.

Even if you don’t give a damn about the technology, you should give a damn about the outcome, the end result of these simulations. Because you’re paying for it. And if you investigate it closely, you’ll see that what you’re paying for has a string of values so complex and far-reaching it’s hard to put a price on. So in the U.S., when it comes to mainstream folks who want to know why in the hell their government just went and dropped bajillions on “sewpercompewters” how can one justify the expense?

One start might be by making it a point of national pride. Of competitiveness. Of dominance. Because whether one agrees with it or not, that’s sometimes the only way to get folks behind something. And from that point, you build to make the science relevant—even for people who aren’t, er, big fans of empirical data.

During the annual Top 500 announcements of the fastest systems in the world, the mainstream media collects around the machines, but in all the talk about speed and “X country did better than Y” they are forgetting why that speed—why that performance differentiation—actually matters. And it’s not just about looking like the leader, there are real benefits.

Supercomputers are hard at work crunching away on grand-scale scientific challenges just as often as they are searching the sub-surface of the earth hunting fossil fuels or signs of impending earthquakes. And they are performing crash tests for the cars we drive, modeling better seats for the children in those cars, and predicting the weather we will be driving in.

Supercomputing is pervasive, but it’s too rare that the public hears anything about it. When there is a big new machine here in the United States that happens to be announced on a really slow news days, mainstream media will jump at the chance to dial up the national labs and marvel at the size of the thing, maybe nip off some generalities about the “scientific” things it will do, but the discussion tends to end there.

This is not to say anyone should expect it to receive the same attention and gloss as the broader consumer technology world. There is no doubt that sparkling arena is more broadly appealing because, after all, it’s tech many of us will use on a daily basis. And as time goes on, there is so much that is new and exciting that the news for that industry creates its own endless carousel of a media ecosystem.

So let’s start here and start easy. And we won’t even talk about why you should care that your tax dollars are funding much of what you see below.

This capability in current generation supercomputers means, among many other things the ability to:

Sequence the entire genome, learn from it, and start down the path to true personalized medicine, then use the data from these genomic records to create indices of illnesses and track them in real-time to better understand and treat conditions.

Predict the weather with increasing accuracy, track that weather across time and space, and draw conclusions not just about the next ten days, but back centuries and simulated in the future for a comprehensive view of a changing climate.

And to the above point, supercomputers are being used to design, model, and test new efficient wind turbines, solar panels, batteries, and other energy storage devices.

And on the subject of wind—as well rain, tornadoes, earthquakes, wildfires, and any manner of ills that can befall mankind, there’s a model and simulation for that.

And with all of these calamities that can befall us at any time, we might want to get the hell of this planet at some point, which means supercomputers are helping us design better rockets and engines, not to mention how they are modeling the solar system, the galaxy, and the beginnings of the universe.

But all of that starts getting too far away from the real world, so back to reality:

And if we can’t get off this earth with our supercomputer-modeled rockets in the future we’re going to need drugs. A lot of drugs. And to create drugs that actually work, we need to simulate molecular interactions to watch how different molecules bind and repel.

And at such a point, if we are still earth bound, we will still need all of our planes and trains and automobiles, which are tested from the component to craft-level across simulations and models of crashes, crushes, and constraints.

And to make these things, whether we like it or not, we are still a fossil fuel-driven economy. Supercomputers are central to reservoir modeling efforts and are a singular source of finding the most lucrative drilling locations around the globe.

So, you get the point.

This was not meant to be an exhaustive list, but rather a glance at what matters—what’s missing from the conversation about technology during news slots and tech chatter at water coolers.

So without the ability to provide numbers that actually register, without the possibility to go through the “things they can do” list we very briefly glossed through here, and no connection to the real world via daily tech news, it comes down to each and every tax payer to learn what these machines do, why they’re important, and why they will continue to be worth funding.

We here at The Next Platform are so used to writing for the people who work with these systems for a living, but with a growing set of readers from outside of supercomputing (i.e. in mid-sized enterprise datacenter environments) it seemed appropriate to level-set. Indeed, much of what the general sys admin at company deals with has little to do with these machines, but there are indeed reasons, both technical and social, that this stuff should matter. And matter to technical folks as well as to “everyday” people who don’t follow tech trends beyond consumer devices.

It’s hard not to be baffled by the lack of understanding about the role supercomputing plays in our daily lives; from the products we use, to the science we (sometimes) accept, to the impact on competitiveness, politics, and hell, even our taxes.

The point is, do yourself a favor if this is all new to you. Go to that same supercomputing resource you probably use daily (Google) and take five minutes to type in your state (or country) and the word supercomputer. You’ll find a center—take a look at their projects or use cases page and see the diversity of what is happening.

You’ll be amazed.

"No, HAL. No more chess. What I really need is a multiphysics simulation across millions of cores. And all you can do is play chess and mess with mind."
“No, HAL. No more chess. What I really need is a multiphysics simulation across millions of cores. And all you can do is play chess and mess with my mind.”

Sign up to our Newsletter

Featuring highlights, analysis, and stories from the week directly from us to your inbox with nothing in between.

Subscribe now

1 Comment

  1. “The fictional stories about supercomputers (Skynet, Mother, HAL, and so forth) rarely depict benevolent systems” … Mycroft Holmes in “The Moon is a Harsh Mistress” (Robert A. Heinlein) seems fairly benevolent (depending on your perspective of course – residents of Earth might have a different view from the Looonies once the rocks started raining down) to me!

Leave a Reply

Your email address will not be published.


*


This site uses Akismet to reduce spam. Learn how your comment data is processed.