Benchmark Space Systems Cuts Simulation Times Over 85% with Rescale and AWS

Benchmark Space Systems develops unique propulsion technology to economically move satellites into more advantageous orbits after being launched into space by larger rockets. The rapidly growing market for its products, known as on-orbit servicing, assembly, and manufacturing (OSAM), is already estimated in the billions. Maintaining a complex ecosystem of satellite infrastructure requires extensive R&D and compute-intensive simulation.

The engineering team uses high performance computing (HPC) resources to run computational fluid dynamics (CFD) simulations before they go to production. To avoid bottlenecks due to limited compute capacity that might lead to delayed launch dates, the company transitioned from running HPC on-premises to AWS. By using Rescale’s HPC automation platform to manage AWS services, Benchmark engineers reduced simulation times from eight hours to one hour, freeing them up to find new ways to grow their offerings and the business. “With revenue growth over 400 percent last year, and five milestone missions set to launch in 2021, we have exceeded our fundraising and growth expectations for the past two years," said Chris Carella, Executive Vice President of Business Development and Strategy for Benchmark Space Systems.

Started in 2017, Benchmark Space Systems is regarded as the go-to for in-space mobility solutions, serving satellite and spacecraft operators in the Space 2.0 segment. It strives to deliver the most cost-effective propulsion solution for small satellite mission needs, from simple orbit adjustments to high-agility maneuvers. The company’s in-space mobility products combine hardware, software, and services to address common operational needs, including last-mile orbit insertion from the rocket. Since its inception, Benchmark has grown 100 percent each year thanks to its ability to get new solutions into orbit in a rapidly developing and competitive market.

Benchmark engineers use computation fluid dynamics (CFD) to model many of its thermal and structural designs. It takes hundreds of thousands to millions of core hours per year to run CFD workloads within a reasonable time frame. In some cases, engineers may oversimplify the models so they can use computer power and speed up time to results, which can lead to errors or inaccuracies.

Because of the mission critical nature of Benchmark’s designs, engineers needed to maintain the full integrity of their tests, but they also needed to deliver on extremely tight timelines. “We anticipated that computing capacity could easily become a bottleneck that prevented us from delivering best-in-class lead times, especially when we’re asked to build first-time configurations for multiple customer programs in parallel,” explained Carella. The company went looking for on-demand access to high performance computing (HPC) resources to run CFD. Additionally, they wanted an automated solution that would manage and optimize the underlying HPC resources for the startup-size team.

Traditionally, engineers use different systems for different activities like running simulations, storing data, building models, sharing ideas, or publishing results. Not only does each system require its own log-in, but engineers also need to manually transfer data and information. With the Rescale HPC automation platform, Benchmark engineers use a single cloud-based control plane, which enables improved data sharing, workflow collaboration, and management control and visibility. By deploying the Rescale HPC automation platform using AWS services, Benchmark can ensure efficient iteration on its model simulations and eliminate islands of analysis, while simultaneously cutting simulation runtimes compared to on-premises HPC systems. Rescale delivers an intelligent computing platform that supports CFD simulations and configures, manages, and optimizes HPC resources automatically for engineers.

On average, it can take eight hours to run a model simulation and optimizing for each mission can sometimes take 20 iterations per system. “That means it takes 20 full days of development iterations to optimize on a complex or sophisticated system configuration,” explained Carella. “And a single system can run up to 16 thrusters.” By using HPC resources on AWS instead of the on-premises system, the team reduced its 8-hour simulation to one hour. Similarly, the Rescale solution abstracts the technology management and free up engineers.

The Rescale HPC automation platform offers a turnkey solution that enables Benchmark’s team to get the best performance, cost, and configuration out of its AWS environment. Built for hybrid and cloud HPC and geared specifically for companies in the R&D phase, Benchmark uses Rescale to further optimize the use of AWS services. “Rescale offers an end-to-end cloud HPC solution that maximizes simulation and modeling processes as well as business management, automation, budget management, and security compliance processes,” said Dylan Collie, Account Executive at Rescale.

Rescale's solution and AWS services offer customers like Benchmark better control and flexibility over how they run simulations even though Rescale selects the server instances and auxiliary HPC resources for them behind the scenes. “Through AWS, we have access to the latest and greatest infrastructure and core types available at this day and age,” said Collie.

During the next 10 years, an estimated 50,000 satellites are projected to launch into space. They’re all going to need specific mobility solutions in order to successfully complete the mission. Running the Rescale platform with AWS services ensures Benchmark can perform quick, efficient model simulations to optimize complex system configurations—helping space customers take one small step for their satellites and one giant leap for their kind of business.