Outsourcing HPC gave Aero Xpertise the compute power and pricing structure it needed to run computational fluid dynamics simulations in parallel for its airline satellite communications customers.
- 83% faster simulations
- Expanded capability to run larger simulations in parallel
- Efficiencies translated to savings passed on to customers
US-based Aero Xpertise specializes in applying computational fluid dynamics (CFD) to solve complex aerodynamic design problems, often in partnership with Aeromechanical Solutions (AMS). Together, the two companies have tackled more than 200 projects, performing thousands of CFD simulations to solve design issues for companies like airline satellite communications (satcom) provider Gogo. Gogo provides next-gen in-flight connectivity solutions for commercial airliners.
CFD simulations in a crunch
Gogo needed to mount its 2Ku satellite antennas and radomes on a wide variety of aircraft, without sacrificing the aircrafts’ aerodynamics, performance or safety. Radomes — the weatherproof enclosures that protect the radio antennas and are effectively transparent to radio waves — are installed on the outside of the aircraft, which means their shape and placement can affect the aerodynamic loads on the structure of the plane.
Will de Jong, Aero Xpertise owner, was tasked with performing CFD simulations to test designs that accounted for all the changing variables. With each one requiring 20 to 40 million data points to model the airplane, not to mention a tight schedule for the client, de Jong needed the computing power and resources of high-performance computing (HPC).
That pricing structure did not compute
Prior to seeking an HPC partner, de Jong ran CFD simulations using his in-house computer cluster. Each simulation took between eight to 12 hours to complete, and without the infrastructure necessary to run simulations in parallel, de Jong was limited in how much he could accomplish in a day. The only way to speed things up was to use fewer data points, which lowered the accuracy of the results. To get the precision he needed, therefore, he spent more time waiting on simulations to finish, leaving less time to analyze and post-process the results.
To keep up with the workload and clients’ demanding schedules, de Jong needed more computing power. However, expanding his in-house cluster would require significant infrastructure upgrades for both power and air conditioning to accommodate more machines. He found it difficult to justify the investments required to expand his cluster, especially when it wouldn’t be in use all the time.
That’s when de Jong started looking at HPC providers. After trying one cloud-based solution he found the system too labor-intensive, and the pricing structure didn’t work well for his type of work. He then discovered the flexibility, ease of use, and simple pricing structure of the Nimbix Cloud and has used it continuously for several years.
HPC as a service keeps computations dynamic
de Jong runs CFD simulations to test aerodynamic designs and the effect of changes on performance. He uses Siemens STAR-CCM+ software with power-on-demand licensing and the Nimbix Cloud, saving time by running simulations faster and running three or four jobs in parallel. The ability to handle larger simulations and run them in parallel creates an efficiency that translates to both a cost and time advantage in his work.
Most of de Jong’s work is with airliner satcom companies like Gogo and with special mission aircraft developers. For both groups, he uses CFD to answer critical questions like how the shape and placement of radomes change the aerodynamic loads on the aircraft structure. This allows him to help his customers improve their designs, answering questions such as where to place the antennas for minimum drag, optimal takeoff performance and maximum fuel economy. And since each airline flies more than one type of aircraft, radomes have to be designed and optimized for all the different airframes.
Faster, higher confidence
By outsourcing HPC to Nimbix, de Jong significantly accelerated his simulations, reducing typical processing times from 12 hours to under two hours. He can also run simulations in parallel.
Because of the Nimbix computing resources, he’s able to improve mesh fidelity and resolve the detailed physics of the airflows, which provides higher confidence in the results, while running more simulations with faster turnaround times. As a result, de Jong has been able to provide these solutions for Gogo:
- Run simulations to evaluate different conditions on baseline aircraft, then run them again with the modification installed.
- Influence the performance of the aircraft, determining where to place antennas for minimum drag and maximum fuel economy.
- Run simulations for any aircraft, allowing airlines to install satcom radomes on any carrier and any aircraft model.
83% drop in simulation time
de Jong leverages the computing resources of the Nimbix Cloud to run CFD simulations faster and in parallel, dropping simulation turnaround time by up to 83 percent. Since the Star CCM+ license is calculated per hour, this in turn reduces the licensing cost.
The additional computing power allows him to increase mesh sizes while still running simulations in a reasonable amount of time. More data points translate to more confidence in the results of the simulations, and shorter simulation times allow him to evaluate additional conditions and provide more data to his clients.
The cost and time efficiency of Nimbix HPC has a direct effect on de Jong’s business and his customers. “It’s more efficient to run larger simulations and run jobs in parallel,” says de Jong. “Frankly, it’s less expensive to pay Nimbix for the additional horsepower because with the lower licensing cost, it ends up reducing the total cost of a simulation. I can lower my prices to my customers, which helps both Aero Xpertise and my customers stay competitive.”
Atos’s Nimbix supercomputing suite is a set of flexible and secure as-a-service high-performance computing solutions. It offers industry-first federated supercomputing-as-a-service and dedicated bare metal services, providing customers with added agility for their compute-intensive workloads and expanded consumption models.