The capabilities of cloud computing resources and more importantly, the price, have increased and decreased respectively in recent years, making their use for computationally complex operations, such as CFD and other simulation tasks, far more accessible. In the past, if a race team wanted to undertake complex CFD simulations, the only option was to invest in an in-house computing cluster.
However, there are now a number of companies offering CFD as a service, harnessing cloud based computing to provide scalability, making aerodynamic simulation accessible to teams for whom it was previously out of reach. On such company is UK based D2H Advanced Technologies, which works with teams across the world, including those in NASCAR. Here, Noah McKay, engineering director at D2H, explains how cloud resources, coupled with efficient CFD workflow tools, allows small outfits, such as JTG Daugherty Racing, a small Cup team with which the company worked in 2020, to compete closer to parity with the big players.
How significant has the move to cloud computing resources been in NASCAR?
I’d say all of the big teams I know of are doing their work in the cloud now. I don’t know if that’s a function of the resources available here in the US, we’ve got a few really big providers here like Amazon Web Services (AWS). Then you’ve got maybe a dozen resellers of AWS, and they’re all competing with each other to be a cloud provider for anybody who need lots of computing hours.
Even as computers and server prices have come down, you can get into a cloud compute system here for much less than you used to. It’s still cheaper to go to the cloud mainly because of the management element, you don’t have to have the staff to manage the systems yourself.
In Formula 1, there is still some hesitancy to harness the cloud, due to rules constraints, but also security concerns, is the latter a consideration?
So far, every group we’ve worked with has kept security at the very front of their operations, and the cloud might actually be more secure than an on-site system. At the end of the day, we’ve got to be linked to the internet no matter what. With respect to Microsoft and AWS, both of those do business for the military and are already equipped with that security infrastructure, which means they can keep up with its changing demands much faster than we would ever be able to.
Are there benefits to being able to scale cloud resources and are there downsides compared to having in-house clusters?
If we’re in a trough, we can run zero [cores], and it cost us nothing, we don’t have any hardware on site that’s costing us money. And when we need that huge bump to say 4000 or 5000 cores, we can have them.
That said, it’s not instantaneous, like it would be if the resource was on site, but it’s not far off though. Even for a big ask, as long as you’re in one of these bigger web services company, you’re only talking maybe five, six hours max, that you’re waiting in the queue. If you’re asking for a 20-hour job and you have to wait an extra five hours for it. That’s no big deal
What other tools do you have available beyond cloud resources
We’ve done some development of our own internal CAD checker software which attaches to your CAD program and puts the tools in the hands of the designer, rather than the aerodynamic engineer. In the past, it’s always been down to the aerodynamic engineer to do the very labour intensive job of getting meshes ready for solution.
That is expensive because you’ve got a heavy-duty aerodynamics engineer spending all of his time just celaning up nasty meshes to a quality good enough to run. We automate those tools that the aerodynamics engineer would have in his head into the CAD program. The designer basically runs a tool that does a number of different checks throughout the design process and says, okay: fix this, fix this and fix this, this won’t work, etc, and highlights everything while telling you how to fix it.
The way our automation system works you just say, here’s my run, here’s the part number I want to change, press go and automation the takes over from there. You know the part is going to pass because we’ve already done the check in CAD. That means the designer can push everything directly through all the way to a solution. That means the aerodynamics engineer can spend all of their time just looking at results.
What are the highlights of your work with JTG?
We’ve had a really good partnership with JTG. They don’t have the level of resources it takes to run a full blown CFD department like a Hendricks Motorsports would. But you know, we’ve got just as much experience, if not more, than the guys we know that are running those departments at teams like Hendrick and Penske.
Our goal is to try to bring that same level of engineering, that same level of expertise, to the smaller teams. From that, they’ve actually picked up a number of spots on the grid. They went in with the attitude of, okay, we’ve got this much budget, how do we get our biggest bang for our buck?
I think as the years go on, we’ll start expanding beyond just the small teams, and more likely into some of the larger teams as they try to reduce staff. Again, for them, aero becomes a service just like we were talking about with crowd computing, they ca turn it on or turn it off as and when they need.