Reducing Simulation Cost to Become More Competitive

Like many aerospace engineering firms, TLG employs STAR-CCM+, a leading industry application, to perform CFD simulations. TLG uses the application to conduct aerodynamic simulations on aircraft and predict the pressure and temperature surrounding airframes. However, the company wanted to reduce the costs associated with running simulations. “We were using a cloud provider to host our simulations, but the cost per simulation was high,” says Andrew McComas, engineering manager at TLG Aerospace. “Running a typical simulation was costing us hundreds of dollars per case, and there may be hundreds of cases per project.”

TLG also wanted the ability to scale its high-performance computing (HPC) applications to take on larger simulations. “The trend in our industry is toward doing more complex simulations that require more compute resources,” McComas says. “But with the internal HPC cluster we were using, we were limited as far as the maximum size problem we could run. We were limited to a small number of nodes and couldn’t allocate enough memory to run large-scale problems.”

Because it wanted to reduce costs and gain scalability, TLG decided to search for a new cloud provider.

TLG uses Amazon Simple Storage Service (Amazon S3) buckets to store multiple terabytes of simulation data on the cloud. The company also takes advantage of Amazon Elastic Block Store (Amazon EBS), which offers persistent block-level storage volumes that can be used with Amazon EC2 instances. TLG uses Amazon EBS volumes for the HPC cluster head node. In addition, TLG uses the Amazon CloudWatch monitoring service to track the progress of CFD simulations in an online AWS console. “We created a CloudWatch metric that enables us to monitor simulation status directly from AWS,” McComas says. “As a result, we can monitor solutions from any computer and are not required to utilize the CFD code’s proprietary GUI.”

• Reduced CFD simulation costs and pass savings to customers to become more cost competitive.

• Access to simulations from all locations, to active monitor jobs and catch problems early to reduce rework.

• Scaibility allows more complex jobs to be processed, resulting in increased manufacturing capabilities.

• 75% reduction in cost per CFD simulation