Cloud Fueled CFD Speeds Up NASCAR Team
By deploying CFD software in a cloud computing environment, NASCAR’s Michael Waltrip racing team is honing their competitive edge using new technology that delivers valuable information in far less time than traditional methods.
A two-time winner of the Daytona 500, Waltrip decided to create his own racing team back in 1996. Today, his team has three Toyota’s competing in the Sprint cup. Driver Clint Bowyer leads the group in fourth place.
Aerodynamic research is important during the design of any vehicle, but it’s essential for success in NASCAR racing. Waltrip’s team designs vehicles that weigh over 3,400 pounds and run around a track at speeds that can exceed 200 mph. Formula One cars also break the 200 mph barrier, but they can weigh half as much as a stock car.
To achieve these high speeds, NASCAR teams are turning to digital simulation software like CDadapco’s STAR CCM+. The application enables race engineers to determine what aerodynamic redesigns are needed to gain a competitive edge.
“NASCAR really is an arms race and digital simulation is one of our best tools to help win that race,” says Andy Hogg in a CDadapco promotional video. A graduate from Duke University’s Master of Engineering Management program, Hogg is now the aerodynamics manager for Michael Waltrip’s racing team.
While Hogg’s use of digital simulation may be new to NASCAR, the use of aerodynamic software has been popular in other areas of motorsport. Speaking with DMR, CDadapco’s David Vaughn explains that a “significant majority” of Formula One racing teams are familiar with STAR CCM+. “Those guys are a little bit further down the road than NASCAR…. it’s a massive engineering effort,” he says.
NASCAR is playing catch up and out of necessity is embracing digital simulation and computational fluid dynamic (CFD) software. For example, next year’s NASCAR vehicles will undergo significant design changes. The bodies will lose their current egg-like shape in favor of styles that look more like actual production vehicles. CFD enables the teams to test upcoming changes without a physical prototype. Just drop a CAD model of a new car and the software does the rest. The digital simulations show how the wheel wells, windshield and other elements affect the car’s performance
CDadapco’s approach to licensing is helping customers like NASCAR adopt the simulation technology. Traditionally, computer aided engineering (CAE) software requires a seat license and a core license, essentially penalizing users with massively parallel clusters.
In response, CDadapco introduced their “power session” model, which allows the use of unlimited cores for one price. They also launched a Power-On-Demand program, which allows customers to use their software in cloud infrastructures. Similar to the power session, CDadapco’s Power-On-Demand does not charge based on how many cores are used – instead, the cloud licensing model only charges by the hour.
Waltrip Racing is running CDadapco’s cloud license using Penguin Computer’s POD (Penguin Computing on Demand) service. POD provides a persistent and secure compute environment that executes jobs directly on the compute nodes' physical, not virtualized, cores for true HPC. The Penguin cloud infrastructure has eight to 10 thousand cores available for use. When the racing team needs to run a simulation, they typically hook up to Penguin’s Salt Lake City datacenter.
Matt Jacobs, senior vice president of corporate development at Penguin, comments that Waltrip’s HPC jobs were quite variable with respect to core count. “Sometimes they come in and use a couple hundred cores, sometimes they’ll use 500,” he says. “It really depends on what they’re trying to get done and how close they are to a race”
Jacobs estimated that purchasing all the hardware and infrastructure to run a 500-core Sandy Bridge cluster would cost roughly $300,000. That number does not include the required power and cooling that goes along with it. Cloud computing is a lot more cost effective.
A major advantage of cloud computing is those customers only pay for what they use. Also, the upfront costs are minimal compared to purchasing a cluster, and end users don’t have to worry about upgrades or maintenance. They also have the advantage of scalability, which enables more complex workloads to be completed faster. Of course, there is a breakpoint where purchasing hardware is financially advantageous to HPC users, but this requires consistent and heavy workloads.
Despite the flexibility and financial incentives offered by cloud computing, the technology does suffer from some issues not associated with local clusters. For example, Jacobs says, users need to communicate with their datacenters via WAN connection, which is not very suitable for moving large amounts of data. There is a point where Penguin recommends its customers simply ship their data on hard drives to populate the service. Given a decent Internet connection, that number is roughly 300 gigabytes.
The Waltrip team is one of those customers with datasets above that threshold. It also had some concerns about data privacy and security, but still decided to run their CFD simulations in the cloud.
As Hogg says, Waltrip racing is determined to win the NASCAR arms race and it plans to do so leveraging high performance computing.
Adds Doug Vaughn, explaining why more companies are turning to these technologies, “It’s all about making simulation on HPC affordable.”