Big Compute - Boom Supersonic – Pushing Boundaries in Aerospace Design

Episode Date: March 14, 2019

Gabriel Broner interviews Josh Krall, the CTO of Boom Supersonic. Boom is designing the next generation of supersonic airliners using HPC that is entirely in the cloud, replacing the decommissioned Concorde and re-opening trans-Atlantic flights at supersonic speeds. For more information, visit www.blog.rescale.com or www.bigcompute.org

Hello, I am Gabriel Bronner, and this is the Big Compute Podcast. Today's topic is pushing the boundaries in aerospace design. We know about Concorde, a supersonic commercial plane that started flying around 1976. It was a technical success, but unfortunately, at around $20,000 a round trip, the economics did not work out, and Cocor stopped flying. More than 40 years later, Boom is embarking in building Overture, a supersonic plane where the promise is that passengers will pay the price of today's business class seats. Boom has been using HPC in the cloud, and in 2017, they received the HPC Wire Reader's Award for Best Use of HPC in the Cloud.

To discuss this new era of supersonic flying, our guest today is Josh Kroll. Josh is co-founder and VP of Technology at Boom Supersonic. Welcome, Josh, to the Big Compute Podcast. Thank you. It's a pleasure to have you here. Welcome, Josh, to the Big Compute Podcast. Thank you. It's a pleasure to have you here. Yeah, thanks. Yeah, let me start by asking you the basic. Why supersonic flying? That's a great question. So I guess I'd start off by saying Boom Supersonic is sort of really

about redefining what it means to fly by reducing travel times by more than half and helping us fulfill the mission of bringing the world closer together. We're the only really, really the only company out there building commercial supersonic aircraft. And you know, our mission is to remove the barriers to experiencing the planet. That's time, money and hassle. We're obviously focused right now on time. And there's sort of been a theme from the beginning of this company around innovation and agility. And that ties to the mission, but also to how we test and validate our aerodynamic designs and use different materials. And also, again, to pushing the boundaries with our computing

needs, which is what we're talking about today. No, that's great. So you say reduce the time in half for all of us who are not thinking too much about this. Give me a sense of what that means if I'm going to cruise the Pacific or the Atlantic. What does it mean for a normal traveler? Right, yeah, it's pretty meaningful from the perspective of a traveler who has an agenda like flying from JFK to London, for instance. We're about 2.6 times faster at Mach 2-2, so roughly cuts travel times in half. And for most of us, we haven't really experienced what cutting travel times in half actually was like in our lifetimes, because the last time we saw that kind of step change in travel was when we went from propeller planes to jets.

But if you look back at history, what we learn is that the second order effects of reducing travel times are just as impactful as the first order. So obviously, for all of us who travel a lot, we'd love to get there faster. That means more time at home with our kids, more time to the work that we care about. But a lot of the second order effects tend to be just as big a deal. The kinds of travel that you choose to do completely change. One of the best examples that I have on that is before the adoption of jet travel, before the introduction of the first jet airliner, and if you wanted to get to Hawaii,

it would take you about 15 hours on a flying boat, a propeller boat, propeller airplane rather. And so not very many people took that trip. And so after the introduction of the first jet airliner, you could make that trip today from the west coast of the U.S. as Hawaii was before the dawn of the jet age. So it gives you a little bit of a sense of what it would do to the world to cut travel times in half. And that's really what we're all about right now. That sounds great. I do remember that when my parents would travel, I'd go to the airport, say bye, it was a big deal. They traveled for a long time. And now we travel perhaps for a week to go to places we can travel for a few hours, I imagine. At time in half, the world. And of course,

if you talk about how it brings people closer together, that means cultures become closer and communication comes closer. So there's a lot of reasons why it's tremendously good for the world. It's great to hear. We're going to feel that the world is a smaller place if this is happening. Let me ask you a question. We know that Concord did fly. So technically, we understand some of the challenges. But because it wasn't viable as a business, it doesn't really succeed over time. So what has changed over time to now make Boom a viable company with a good business plan? Yeah, that's a great question. A lot of people go to Concord and ask that question because obviously we know Concord was successful from a technological standpoint.

In fact, it was in many respects ahead of its time, kind of a technological marvel. But they only made 14 of them and they're only in museums today. So we're working hard to sort of revive that dream and bring that back. I think the fundamental drawback to Concord, as you mentioned, was the efficiency and economics. To start off, it had more seats that could be filled at the ticket prices that were required, and they had to charge those high ticket prices

because of the operating maintenance costs. Concorde was designed at a time when they didn't have access to the today's modern carbon fiber composite materials. They designed using wind tunnels over you know decades of design cycles rather than having today's sort of computational fluid dynamics approaches. And then the propulsion system was a turbojet engine with an afterburner. And, of course, today we have modern turbofan engines, and they're tremendously more efficient than what Concorde was able to use. And so you combine sort of those 50 years of advances in aerospace technology,

and you can kind of build on the legacy of Concorde. And that allows us to bring Overture to market, which is our sort of first supersonic airliner design. Overture addresses a much larger market as a result of having those efficiencies, and it will be significantly less expensive to operate. So we have 55 seats to fill instead of 100, and operating costs that are similar to subsonic business class. So it's viable on sort of hundreds to thousands of long-haul routes,

and it will be quieter, more fuel-efficient, and more affordable than Concorde, which is why we've been able to kind of make traction with Overture with airlines such as Japan Airlines and Virgin Group, and we've pre-sold 30 of them so far. That's great to hear. So if I summarize in my head, there's new carbon fiber composites, new CFD, new propulsion systems. Those have all helped in terms of enabling the design. new CFD, new proportion systems.

Those have all helped in terms of enabling the design, but also you designing for 55 seats, which gives you a better price. Is that a fair summary of why this is viable? Yeah, and on top of that, I think the other aspect that people don't realize is that the market for overland supersonic travel, or rather business class travel over land or over water sorry i've said over over land i meant over water um long haul business class travel market has increased dramatically since the days when concord was

entering service and so today there's uh just a dramatically larger market and that that helps with the operating economics because you need the economies of scale of, you know, hundreds of airplanes and a robust maintenance network in order to bring down the maintenance costs. That's excellent. So this is very good. All the changes have enabled this new innovation. So it's great to hear.

So you went off to design the next supersonic plane and you decided to use HPC in the cloud instead of on-premise HPC which has been more traditional. Can you talk to us about the choice you made? Yeah absolutely. I guess it's worth mentioning that what we've started with as a company is we're building a supersonic demonstrator aircraft so a lot of our work to date especially with our HPC simulation work has been focused on that demonstrator it's proving out the same technologies that we plan to bring to market in the Overture design it's called XB1 and the test the data from

the test flights will sort of be will help us design and refine the engineering of our 55-seat over-the-air airliner. In the earliest days, we were beginning the design of our demonstrator, and we were a startup, and so we had some significant trades between capital investments in on-premises infrastructure and being able to use that capital for other things growing the business. And so, you know, I think that the one aspect of, you know, what Rescale offers and what HPC in the cloud offers is just being able to avoid those upfront infrastructure costs. So that was probably the most meaningful reason early on. Today, of course, we're not as capitally constrained. We have, you know, we've raised

a bunch of money and we have the ability to invest in on-premises infrastructure, but we actually value the cloud for different reasons. And it has to do more with being capacity constrained if you're on-premises. So next month, we're about – well, actually, it starts this month. We're going to run about 13 million core hours of simulation work, which is, as you know, thousands and thousands of machines. And the ability to flex into the cloud and get that work done quickly over days or weeks rather than weeks and months gives us a lot more schedule agility and allows the engineers to get their results when they need them

rather than have to very carefully plan out the high utilization of on-premises infrastructure that we're invested in. So between those two things, I think that's sort of the reason that we tend to use the cloud with most of our simulation work. When you started, you say, well, you're a startup. You cannot afford to go buy a big HPC system. That was the primary reason. But today, the needs for a big simulation in spikes is what motivates you

to continue to use the cloud because of the elasticity. I think that's interesting. At two different points, you have two different reasons to choose the cloud because of the elasticity. So I think that's interesting. At two different points, you have two different reasons to choose the cloud. Honestly, when I joined Rescale in 2017, I was wondering if something like Boom could be designed in the cloud because my background has been mostly on-premises.

And I was pleasantly surprised to see this is happening. What has been your experience in terms of using the cloud? What kind of applications you're in? What kind of results are you achieving with that? Sure. So we have two different teams here that do simulation, period, whether it's in the cloud or not. There's our main external aerodynamics team, obviously,

that's doing full aircraft simulation. These days they're doing pretty detailed database work trying to prove out the flight handling characteristics, stability control, loads, all of these different workflows. They're producing some fairly large data sets for the rest of the group. And then there's our propulsion team that is doing inlet aerodynamics, and they're doing both design work and more complex work like unsteady cases and flutter cases and shock simulations and such. So we have both workflows. We use a lot of different codes here. We started off with some sort of either open source or NASA-based codes like Fun3D and Cart3D.

And we've more recently adopted some other proprietary codes like Numeca and CFD++. So we sort of have, you know, what's the best tool for the job kind of approach, and we have different pools of expertise within our aerodynamics and propulsion teams who have their tool of choice. As far as results, I mean, we're modeling things like complex vortex flow over a delta wing at low speed conditions, which is not an easy problem to solve same thing with inlet aerodynamics i mean the concord designers took about a decade to get their supersonic inlet design correct and uh and our designers were able to do it in about nine months based on uh you know running the kinds of cft simulations they can do today. We've been validating this against wind tunnel results,

both supersonic inlet wind tunnel and low-speed aerodynamics results from the wind tunnel and getting really good matches. So overall, I'd say we're using CFD the same way any other company is, except that we're running it in a different place. And that allows us to have a lot of flexibility about how we get those results and when. This is great to hear. You're able to compete with large companies developing airplanes.

And you mentioned a couple of airlines have bought your airplanes. Is that right? That's right. Yeah. Virgin and JAL. Okay. When do we expect to fly them? We want to go. Yeah. So we have a lot of steps ahead of us between now and then. We're, we're, most of our team right now is focused on our demonstrator. We're, you know, we're aiming to bring that in, into flight tests by the end of the year and, and supersonic flight tests in early next year. So, so that's the, that's what, you know, the majority of our team is working on right now, but we also have a team doing early designs on Overture and,

and, you know, the timelines on that are, are, are longer term. You know, we, we talk about mid 2020s, but it, there's a lot of steps we hit between here and there. and so um you know uh one step at a time and uh well you tell me when we can fly it then we'll be there exactly listen josh it's great to hear how you have this great vision in front of you uh seems to probably spend lots of nights trying to make it happen but seeing the success in the um the early design happening in airlines buying this. Let me ask you, what challenges do you have in front of you right now? Yeah, I mean, I think from an HPC point of view, the biggest challenge we have is how to manage the growth of the program and the compute demands that it will require as we, as we begin to scale into the overture program over the next year.

We need really high performance and some, some of the, you know, legacy cloud platforms are, are sometimes struggle with that. And so we're, we're working with the rescale team and with those cloud providers on, on trying to get, you know, be able to run these really large billion node kind of problems in the cloud. I think that's kind of pushing the frontiers of what people have done in the cloud in the past. So those are some challenges that I think we're collectively focused on.

And then I think, you know, beyond that, it's just continually looking for how to eke out more performance from these cloud platforms because we're getting tremendous advantage out of it, but we're always balancing that trade between what does on-premises look like. And so far, that has been a pretty easy set of decisions. But as we grow, that can going to continue to be a challenge for us. So far, it's working well. As you increase the size of your simulation, you wonder how will everything respond? And will the demands be there and be met? That's great to hear. So you're ahead

of many people in terms of what you're doing. You're pushing the envelope in aerospace design, but you're also pushing the envelope in terms of moving workloads to the cloud. I think it'd be interesting if you have any learnings, any experience you'd like to share. Imagine I'm a practitioner trying to figure out, should I go to the cloud? Should I not? What works? What doesn't? Anything you'd like to share for everyone else listening today? Sure. I think for the one-off kind of design cases or single engineer kind of workflows, the cloud's ready to go. I mean, you're generally not going to be doing the volume of work

where you need to invest in tooling significantly or be too concerned. I mean, what Rescale offers out of the box is very capable of doing work in the cloud that you would normally do in an on-premises solution. But as you scale up into very large databases where you're running thousands of cases at a time and pulling back terabytes of data and such. I think just like on-premises, you have to invest in that tooling and some of the challenges that you run into with the cloud are slightly different than the kinds of problems you end

up with on-premises infrastructure. So you end up grappling with the fact that cloud is multi-tenant and that your infrastructure is not always as stable, even though you're getting this benefit of being able to burst into a lot more capacity. And so if you're starting to do very large databases, you need to have some tooling infrastructure in place to handle that. That is easy to underestimate. But I think our partnership with Rescale has been really invaluable on that. We've done a lot of co-development where we've built things on our side and Rescale's built things on their side.

And so I think the good news is that we're hoping to pave some ground here, I think, but that is the only thing to be aware of is that as you scale up in the cloud, it isn't yet quite as push-button as you might like. But good progress is being made. Yeah, that's great. I remember my days when I was working on-premise with NASA, and we had the same issues. They were pushing the boundaries, simulating reentry and things like that. And we had to push the boundaries and we had to fix issues as they come up,

but they paved the way for normal users coming after that. So in a sense, you're taking a cost that is more of a service to the rest of the HPC community because you're going after millions and billions of core hours. So you're doing it first. You're paving the way for the rest of the people to make it happen. It's great to hear your story,

how you're pushing the vision of supersonic travel, changing the world. We all want to fly, understanding how you went about choosing HPC in the cloud and what's working, what's not. I'm sure we're going to talk more at some point as things progress in your company. But before we close, anything you'd like to add?

You know, just to reemphasize that what we're working on here is sort of, in some senses, pretty innovative and on the bleeding edge of how to solve these kinds of problems and how to build a new company that can tackle what, you know, in the past, I think only was perceived as being accessible to big aerospace. And so, you know, if that's interesting to people and you'd like to join our team here, I'd love to, you know, ask you to take a look and take a look at our website and see what opportunities are available. We've got a couple of dozen of some of dozen of some of the world-class aerodynamicists

and propulsion designers and software engineers here. And I'd love to talk to talented people who are inspired by this kind of mission. That sounds great. Hope this helps you recruit more talent to continue to grow your company. This is great to hear. So this has been a fascinating story and I would like to thank our guest, Josh Kroll,

co-founder and VP of Technology at Boom Supersonic for pushing the boundaries in the area of supersonic travel and for sharing his experience that we may apply in our areas. Till next time, I am Gabriel Bronner

and this was The Big Compute Podcast.