Democratize Access to HPC for Computer-Aided Materials Design Using Amazon EC2 Spot Instances with Good Chemistry

Founded in 2021, Good Chemistry has a mission to create a more sustainable, circular economy by solving tough material science problems, like the removal of PFAS from the environment. Its product, QEMIST Cloud, uses high performance computing (HPC) clusters on AWS to push the boundaries of what is possible with quantum chemistry simulations. Using these simulations, scientists can accelerate the discovery and development of new materials. “The number of potential synthesizable molecules dwarfs the number of particles in the observable universe,” says Zaribafiyan. “Our mission is to use modern computing on the cloud to search uncharted chemical space and bring new materials and new drugs to market faster.”

Finding affordable, scalable ways to break the chemical bonds in PFAS is a major priority for scientists around the world. These artificial chemicals are found in everything from nonstick cookware to firefighting equipment but are known to cause significant health problems, including harm to the reproductive and immune systems and an increased risk of cancer. “Because PFAS are not biodegradable, they accumulate in the environment and find their way into underground water reservoirs,” says Zaribafiyan. “In the United States alone, more than 200 million people have PFAS in their drinking water. That’s two-thirds of the population.”

To speed up chemistry simulations in the cloud, Good Chemistry joined forces with the AWS HPC team and Intel as part of the Amazon Global Impact Computing team’s initiative on Digital Technologies for a Circular Economy. Together, the teams developed highly scalable infrastructure for QEMIST Cloud powered by AWS services like Amazon Elastic Compute Cloud (Amazon EC2) Spot Instances, which run hyperscale workloads at significant cost savings. Through this engagement, Good Chemistry massively increased the scaling capabilities of QEMIST Cloud to run a chemistry simulation using more than one million CPU cores.

On AWS, Good Chemistry can run high-throughput, high-accuracy HPC workloads at scale. QEMIST Cloud’s infrastructure is containerized and uses Amazon Elastic Kubernetes Service (Amazon EKS) to start, run, and scale Kubernetes clusters, each of which runs chemistry algorithms. Using Karpenter, an open-source node provisioning solution, each HPC cluster can scale on multiple instance types across Availability Zones, providing optimal scale and availability. “Using this approach, we can take advantage of all Availability Zones in an AWS Region and circumvent any scaling issues that Kubernetes might encounter,” says Rudi Plesch, head of software development at Good Chemistry. “Periodically, we rebalance the clusters to make sure that none of them run out of work.” The immediate results of the simulation are then stored on Amazon Aurora, a relational database management system built for the cloud with full MySQL and PostgreSQL compatibility.

Using its highly scalable AWS infrastructure, Good Chemistry accurately calculated the bond-breaking energy for PFOA, one of the largest and most notorious PFAS molecules, in 37 hours, with only 4 hours at the one million core peak. Had the company tried to run these simulations sequentially, the process would have taken several years. “We dynamically scaled QEMIST Cloud to one million cores, and by the next day, we were able to create a new solution that was out of reach before,” says Zaribafiyan. “All it took was the on-demand scalability of the cloud. It’s a game changer for HPC in material science and chemistry.”

With QEMIST Cloud, Good Chemistry has democratized access to supercomputer capabilities for research organizations, regardless of size or resources. “You don’t have to spend millions of dollars in infrastructure to get computing capability at this scale,” says Philip Ifrah, head of product at Good Chemistry. “Our solution on AWS orchestrates millions of computing resources on demand to perform experiments that push the boundaries of what’s possible.”