<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=1875728222659097&amp;ev=PageView&amp;noscript=1">

ONERA M6 Wing Benchmark on 2 HPC Clusters

Home » Blog » ONERA M6 Wing Benchmark on 2 HPC Clusters

One of our European trial users wanted to compare the performance of EXN/Aero with the Computational Fluid Dynamics solver they are currently using - SU2. Their company is an Engineering Consulting firm that provides design services to the aerospace sector. They’re involved in cutting edge aircraft designs and rely heavily on tools such as CFD to deliver accurate results to customers. Like any consulting firm, they’re mindful of cash, but they also want to provide the highest quality results to clients in a timely manner.

The case we used for the benchmark was the ONERA M6 Wing. This geometry is based on a scale wind tunnel model and is referenced by NASA and ONERA as a validation case for CFD software with compressible / transonic flow capability — in particular their ability to resolve normal shocks on the surface of the wing. According to documentation on the ONERA website, the model and range of experimental conditions are meant to reflect conditions seen in both military and civil aviation scenarios. Full details about the geometry, domain, boundary conditions, and mesh resolution can be found here

This is a 9M cell structured multiblock mesh that was run on our HPC cluster and their HPC Cluster. The same mesh was used in both cases, and the simulation was run for two ‘wash throughs’ of initial conditions in order to obtain accurate surface pressure values. Below is a view of the pressure coefficient distribution across the wing span for different Mach numbers (taken from the EXN/Aero simulation results file). 

On the user's HPC Cluster, they had access to 160 CPUs for this simulation. Using SU2, the simulation took 200 hours to reach accurate pressure values at a cost of $8.80 / System Hour. On our HPC Cloud, we used 3 CPUs and 2 Nvidia K80 GPU cards. Using EXN/Aero, the simulation took 67 hours to reach accurate pressure values at a cost of $4.85 / System Hour. The total compute cost using SU2 was $1,760 and using EXN/Aero was $324. 

Solver

SU2

EXN/Aero

Simulation Time (hours)

200

67

Compute Rate (USD)

$8.80 (160 CPUs)

$4.85 (2 K80 GPUs + 3 CPU)

Compute Cost (USD)

$1,760

$324

 

What does this mean for CFD engineers?

CFD Engineers are under immense pressure to deliver more simulation output, higher resolution results, larger & more complex models, and to run simulations on new applications. They're expected to do this while minimizing cost, minimizing resource usage, and operating on 30 year old license models. 

This article's ONERA Wing benchmark first demonstrates how EXN/Aero is innovative in the way that it reduces simulation times and expense. This enables engineers to run larger, higher resolution, and more complex simulations for a lot less cost. The second way it is innovative is how you can access the solver on-demand via our cloud host, a pay-as-you-go model (on top of a nominal monthly subscription fee).

The performance improvement and on-demand ability enable you to take your simulation productivity to the next level, while respecting your cash-flow. This is especially important for engineering consultants who need to manage their productivity and budget at all times. 

Faster solution times + an on-demand model gives you the ability to: 

  • Upsell existing clients on larger or more complex models
  • Offer a higher through-put of simulations to prospective clients
  • Offer shorter turn-around times on simulations
  • Upsell existing customers on high resolution models (DES/LES)
  • Supplement existing CFD workflow using the on-demand tool for additional or more challenging runs
  • Bid more competitively on projects
  • Bid on larger simulation projects outside your compute resource capacity
  • Provision compute resources on the fly to scale with customer demands
  • Test out simulations on new applications 
  • Increase their simulation throughput without purchasing additional license seats
  • Avoid the maintenance fees and overhead expenses associated with managing a cluster

Combining a powerful on-demand CFD solver with the ability to access it on the cloud and the ability to pay only for what you use enables you to compete, grow, and better manage your cash-flow. To get a free trial for a sense of the interface, workflow, and speed up, click here.


We also just joined forces with Pointwise to offer their Meshing Tool on our on-demand platform. Now you can mesh, solve, and post-process using only your browser and our on-demand HPC cluster. 

2017-05-12