The America’s Cup Dream
The 36th America’s Cup race will be decided in Auckland, New Zealand in 2021. Like all the teams, INEOS TEAM UK will compete in a boat whose design will have followed guidelines set by race organizers to ensure the crew’s sailing skills are fully tested.
A Technology Boat Race
Despite the restrictions, teams still have control over features such as the shape of the boat’s monohull and foils, but with limited on-water testing, engineers must turn to computer-based simulations to optimize their designs. They depend on the computational power available to process thousands of simulations, exploring possible boat shapes and positions on the water. In the case of INEOS TEAM UK, for example, it needs 2,000−3,000 computational fluid dynamics (CFD) simulations to design the dimensions of just a single boat foil.
To run these simulations using the team’s on-premises high performance computing (HPC) resources could take more than a month. Nick Holroyd, head of design at INEOS TEAM UK, says, “With so many design decisions to make before the competition, a month was too long. It reduced the time our engineers had to consider the results, limiting the freedom they needed to be innovative and make the right choices.”
Greater HPC Scale, Lower Cost
The team turned to Amazon Web Services (AWS) to migrate its CFD simulations to the AWS Cloud. The team chose AWS because of the scale of its HPC resources as well as its cost-effectiveness. INEOS TEAM UK could keep its costs low by using Amazon Elastic Compute Cloud (Amazon EC2) Spot Instances, which allow customers to access unused Amazon EC2 capacity.
To get the performance INEOS TEAM UK required and on budget, the team worked with AWS, who helped design an HPC environment based on multiple Availability Zones in multiple regions and Amazon EC2 Spot Instances, which provided a 65 percent cost saving compared to on-demand capacity.
For the hull, whose design needed hundreds of compute cores for every simulation, the team used Amazon EC2 C5 instances in addition to the latest Amazon EC2 C5n Nitro-powered instances with Elastic Fabric Adapter (EFA) network interfaces.
To ensure fast disk performance for the thousands of simulations completed each week, the team also used Amazon FSx for Lustre to provide a fast, scalable, and secure high-performance file system based on Amazon Simple Storage Service (Amazon S3).
“The speed combined with the low cost of the Amazon EC2 Spot Instances means we can do many thousands more simulations within our design budget,” says Holroyd. “One question I constantly ask myself is whether we’re spending our money wisely. Using AWS, I have no doubts because it massively compresses the computational turnaround, maximizing design time.”
Get started with running your CAE/CFD workloads now – fill the form and get a $100 AWS credit!