 From around the globe, it's theCUBE with digital coverage of Exascale Day, made possible by Hewlett-Packard Enterprise. Welcome everyone to theCUBE's celebration of Exascale Day. Dr. Mark Fernandez is here. He's the HPC Technology Officer for the Americas at Hewlett-Packard Enterprise, and he's a developer of the Spaceborne Computer, which we're going to talk about today. Mark, welcome, it's great to see you. Great to be here, thanks for having me. You're very welcome. Let's start with Exascale Day. It's on 10.18, of course, which is 10 to the power of 18. That's a one followed by 18 zeros. I joke all the time. It takes six commas to write out that number. But Mark, why don't we start? What's the significance of that number? So it's a very large number. And in general, we've been marking the progress of our computational capabilities in thousands. So Exascale is a thousand times faster than where we are today. We're in an era today called the petaflop era, which is 10 to the 15th. And prior to that, we were in the teraflop era, which is 10 to the 12th. I can kind of understand a 10 to the 12th, and I kind of can discuss that with folks because that's a trillion of something. And we know a lot of things that are in trillions, like our national debt, for example. But a billion billion is the Exascale, and it will give us a thousand times more computational capability than we have in general today. Yeah, so when you think about going from tera-scale to pedascale to exascale, I mean, we're not talking about orders of magnitude. We're talking about a much more substantial improvement. And that's part of the reason why it sort of takes so long to achieve these milestones. I mean, it kind of started back when they were in the 60s and 70s. And then we've been in the pedascale now for more than a decade, if I think I'm correct. Yeah, correct. We got there in 2007, and each of these increments is an extra comma. That's the way to remember it. So we want to add an extra comma and get to the exascale error. So yeah, like you said, we entered the current pedaflop scale in 2007. Before that was the tera-scale, tera-flop error, and it was in 1997. So it took us 10 years to get that far, but it's taken us, kind of take us 13 or 14 years to get to the next one. And we say flops. We're talking about floating point operations. And we're talking about the number of calculations that can be done in a second. I mean, you talk about not being able to eat your head around it, right? Is that's what we're talking about here? Correct. Scientists, engineers, weather forecasters, others use real numbers and real math. And that's how you want to rank those performance. It's based upon those real numbers times each other. And so that's why they're floating point numbers. When I think about supercomputers, I can't help but remember whom I consider the father of supercomputing, Seymour Cray. Cray, of course, is a company that Hewlett-Packard Enterprise acquired. And he was kind of an eclectic fellow. I mean, maybe that's unfair, but he was an interesting dude, but very committed to his goal of really building the world's fastest computers. When you look back on the industry, how do you think about its developments over the years? So one of the events that stands out in my mind is I was working for the Naval Research Lab outside Stennis Space Center in Mississippi, and we were doing weather modeling. And we got a Cray supercomputer, and there was a party when we were able to run a two-week prediction in under two weeks. So the scientists and engineers had the math to solve the problem, but the current computers would take longer than just sitting and waiting and looking out the window to see what the weather was like. So when we can make a two-week prediction in under two weeks, there was a celebration. And that was in the 80s, early 90s. And so now you see that we get weather predictions in eight hours, four hours, and your morning folks will get you down to an hour. Well, so, I mean, if you think about the history of supercomputing, it's really striking to consider the challenges and the efforts as we were just talking about. I mean, decade plus to get to the next level. And you see this coming to fruition now, and we're saying Exascale likely 2021. So what are some of the innovations in science and medicine or other areas? You mentioned weather, that'll be introduced as Exascale computing is ushered in. What should people expect? So we kind of alluded to one and weather affects everybody everywhere. So we can get better weather predictions, which help everybody every morning before you get ready to go to work or travel or et cetera. And again, storm predictions, hurricane predictions, flood predictions, the forest fire predictions, those type things affect everybody every day. Those will get improved with Exascale. In terms of medicine, we're able to take, excuse me, we're able to take genetic information and attempt to map that to more drugs quicker than we have in the past. So we'll be able to have drug discovery happening much faster with an Exascale system out there. And to some extent, that's happening now with COVID and all the work that we're doing now. And we can, we realize that we're struggling with these current computers to find these solutions as fast as everyone wants them. And Exascale computers will help us get there much faster in the future in terms of medicine. Well, and of course, as you apply machine intelligence and AI and machine learning, to the applications running on these supercomputers, that just takes it to another level. I mean, people used to joke about, you can't predict the weather. And clearly we've seen that get much, much better. Now, it's going to be interesting to see with climate change, that's another sort of wild card variable. But I'm assuming the scientists that are taking that into consideration. I mean, I've actually been pretty accurate about the impacts of climate change, haven't they? Yeah, absolutely. And the climate change models will get better with Exascale computers too. And hopefully we'll be able to build a confidence in the public and the politicians in those results with these better, more powerful computers. Yeah, let's hope so. Now let's talk about the space-born computer and your involvement in that project. Your original space-born computer, it went up on a SpaceX reusable rocket. The destination, of course, was the International Space Station. Okay, so why, what was the genesis of that project and what was the outcome? So we were approached by a longtime customer, NASA Ames. And NASA Ames has its mission is to model rocket launches and space missions and return to Earth. And they had the foresight to realize that their supercomputers here on Earth could not do that mission when we got to Mars. And so they wanted to plan ahead and they said, can you take a small part of our supercomputer today and just prove that it can work in space? And if it can't, figure out what we need to do to make it work, et cetera. So that's what we did. We took identical hardware that's President NASA Ames. We put it on a SpaceX rocket. No special preparations for it in terms of hardware or anything of that sort. No special hardening because we want to take the latest technology just before we head to Mars with us. I tell people, you wouldn't want to get in the rocket headed to Mars with a flip phone. You want to take the latest iPhone, right? And all of the computers on board current spacecrafts are about the 2007 error that we were talking about in that error. So we want to take something new with this. We got the spaceborne computer on board. It was installed in the ceiling because in space there's no gravity and you can put computers in the ceiling. And we immediately made a computer run and we produced a trillion calculations a second. Which got us into the terraflop range. The first terraflop in space was pretty exciting. Well, that's awesome. So this is the ultimate example of edge computing. You mentioned you wanted to see if it could work. And it sounds like it did. I mean, there was obviously a long elapsed time to get it up and running because you have to get it up there. But it sounds like once you did, it was up and running very quickly. So it did work. What were some of the challenges that you encountered? Maybe some of the learnings in terms of getting it up and running? So it's really fascinating. Astronauts are really cool people, but they're not computer scientists, right? So they see a core. They see a place to plug it in. They plug it in. And of course we're watching live on the video and you plugged it in the wrong spot. So Mr. Astronaut, can we back up and follow the procedure more carefully and get this thing plugged in carefully? They're not computer technicians used to installing a supercomputer. So we were able to get the system packaged for the shake, rattle and roll and G-forces of launch in the SpaceX. We were able to give astronaut instructions on how to install it and get it going. And we were able to operate it here from Earth and get some pretty exciting results. So our supercomputers are so easy to install. Even an astronaut can do it. I don't know. Correct. Here on Earth we have what we call customer replaceable units and we had to replace a component. And we looked at our instructions that are tried and true here on Earth for average Joe, a customer to do that and realized without gravity, we're gonna have to update this procedure. And so we renamed it an astronaut replaceable unit and it worked just fine. Yeah, you can't really send an SE out to space to fix it. No, sir. Get to have very careful instructions for these guys but they're great, worked out wonderfully. That's awesome. Let's talk about Spaceborne 2. Now that's on schedule to go back to the ISS next year. What are you trying to accomplish this time? So in retrospect, Spaceborne 1 was a proof of concept. Can we package it up to fit on SpaceX? Can we get the astronauts to install it and can we operate it from Earth? And if so, how long will it last and do we get the right answers? 100% mission success on that. Now, Spaceborne 2 is we're gonna release it to the community of scientists, engineers and space explorers and say, hey, this thing is rock solid, it's proven. Come use it to improve your edge computing. And A, we want to reserve and we'd like to preserve the network downlink bandwidth for all that imagery, all that genetic data, all that other data and process it on the edge as the whole world is moving to now. Don't move the data, let's compute at the edge and that's what we're going to do with Spaceborne 2. And so what's your expectation for how long the project is gonna last? What does success look like in your mind? So Spaceborne 1 was given a one year mission just to see if we could do it, but the idea then was planted, it's gonna take about three years to get to Mars and back. So if you're successful, let's see if this computer can last three years. And so we're going up February 1st if we go on schedule and we'll be up two to three years. And as long as it works, we'll keep computing and computing on the edge. That's amazing. I mean, I feel like, when I started the industry, it was almost like there was a renaissance in supercomputing. You certainly had Cray and you had all these other companies. You remember thinking machines and convex spun out, try to do a mini supercomputer. And you had really a lot of venture capital and things got quiet for a while. I feel like now with all this big data and AI, we're seeing in all the use cases that you talked about, we're seeing another renaissance in supercomputing. I wonder if you could give us your final thoughts. Yeah, absolutely. So we've got the generic, like you said, floating point operations. We've now got specialized image processing processors and we have specialized graphics processing units, GPUs. So all of the scientists and engineers are looking at these specialized components and bringing them together to solve their missions at the edge faster than ever before. So this heterogeneity of computing is coming together to make humanity a better place. And how are you going to celebrate Exascale Day? You got a special cocktail, you going to shake up what are you going to do? It's five o'clock somewhere on 10, 18 and I'm a parrot head fan. So I'll probably have a Margarita. There you go. All right. Well, Mark, thanks so much for sharing your thoughts on Exascale Day. Congratulations on your next project, the Spaceborne 2. Really appreciate you coming to theCUBE. Thank you very much. I've enjoyed it. All right, you're really welcome. And thank you for watching everybody. Keep it right there. This is Dave Vellante for theCUBE. We're celebrating Exascale Day. We'll be right back.