For over four decades, the vital quantum chemistry software GAMESS has empowered more than 140,000 researchers globally, but accessibility has been a growing challenge. Now, Old Dominion University (ODU) is collaborating with Iowa State University and Amazon Web Services (AWS) to modernize the two-million line code package, creating CPU– and GPU-optimized containers deployable on AWS High-Performance Computing services. This move aims to standardize computational environments, boosting reproducibility and collaboration for scientists modeling everything from new materials to potential drugs. “Moving to the cloud allows us to balance performance and power consumption in an environmentally conscious way,” explains ODU professor Masha Sosonkina, highlighting the project’s commitment to sustainable scientific advancement. This partnership offers a blueprint for bringing legacy high-performance computing applications into a modern, cloud-native future.
GAMESS Software Modernization Addresses Reproducibility Challenges
A critical overhaul of the GAMESS quantum chemistry software is underway, aiming to resolve longstanding issues with reproducibility and accessibility for its 140,000 global users. This isn’t simply a port; it’s a fundamental shift addressing the inconsistencies arising from researchers utilizing vastly different computational setups, ranging from supercomputers to personal laptops. The core problem stems from the difficulty in achieving consistent results across these disparate environments. As Peng Xu of Iowa State explained, “When we do a GAMESS calculation on one system, it needs to be reproducible no matter what kind of architecture or hardware you run the calculations on.
That’s a scientific requirement.” To tackle this, the team focused on creating CPU and GPU containers, meticulously reconciling versions of compilation tools to ensure consistent performance. The workflow now leverages AWS Parallel Computing Service (AWS PCS) with Slurm, allowing researchers to utilize familiar commands on Amazon Elastic Compute Cloud (Amazon EC2) instances. Amazon Elastic File System (Amazon EFS) provides shared storage across compute nodes, further streamlining collaboration. Rigorous validation was paramount; the team ran tests on 48 protein-ligand conformers, directly comparing outputs with previously published research.
Using Tuning and Analysis Utilities (TAU) performance analysis software, they confirmed the containers introduced “no performance overhead compared to traditional installations.” This assurance of accuracy is vital for a scientific tool where data consistency is non-negotiable. They are also investigating integrating artificial intelligence to refine molecular conformation selection, potentially accelerating drug discovery. Xu envisions, “We hope that GAMESS can be adapted to an AI-based selection strategy, so it’s not limited to the current hardware and environment.”
CPU & GPU Containerization with AWS HPC Services
This variation historically hampered collaborative efforts and limited participation in advanced computational chemistry. The team constructed both CPU- and GPU-optimized GAMESS containers, a decision driven by the need to offer researchers flexibility based on their specific computational demands. “Matching versions of the tools used to compile GAMESS was our biggest challenge,” said Masha Sosonkina, professor of electrical and computer engineering at ODU, highlighting the complexities of ensuring consistent performance across diverse environments. This meticulous work culminated in a functional, containerized version of GAMESS achieved by August 2025, following an initial AWS workshop in June 2025.
A key aspect of the modernization was validating the containerized GAMESS’s accuracy. This validation was deemed “essential for any scientific application migrating to the cloud,” ensuring researchers could trust the results generated by the cloud-native version. Looking forward, ODU plans to deploy an ARM-optimized container on AWS Graviton Processors, potentially improving energy efficiency by 30 percent with only a 10 percent reduction in computational speed.
Validation of Containerized GAMESS Maintains Accuracy
Rigorous testing confirmed that encapsulating the GAMESS quantum chemistry software within containers didn’t compromise its scientific validity, a crucial step in modernizing the widely-used program. The team deployed a four-node system and directly compared outputs, ensuring consistency with existing data—a vital prerequisite for any scientific software migration. This validation wasn’t merely about matching numbers; it addressed a fundamental challenge in computational chemistry. This is significant because maintaining speed is paramount when dealing with complex molecular simulations. The team’s approach provides assurance to the 140,000 researchers globally who rely on GAMESS for modeling molecular structure and behavior.
Beyond accuracy, the project prioritized efficiency. Researchers are increasingly focused on balancing computational power with environmental impact, and the cloud offers unique advantages. The validated containers are also poised for scalability testing using Elastic Fabric Adapter (EFA) networking, paving the way for handling even more demanding simulations. Looking further ahead, the team intends to integrate artificial intelligence to refine the selection of molecular conformations, potentially accelerating drug discovery.
When we do a GAMESS calculation on one system, it needs to be reproducible no matter what kind of architecture or hardware you run the calculations on. That’s a scientific requirement.
Hybrid Workflows & Future AI Integration for Drug Discovery
A longstanding challenge in computational chemistry – ensuring consistent results across diverse computing environments – is being addressed through a collaborative modernization of the GAMESS software package. Originally developed over four decades ago, GAMESS supports over 140,000 researchers globally, aiding in areas from materials science to crucial drug discovery work like protein-ligand binding predictions. The impetus for this shift stems from the varied infrastructure researchers employ, ranging from supercomputers to personal laptops. This inconsistency impacts the reliability of results and hinders collaborative efforts.
This move towards a hybrid workflow – combining the speed of molecular mechanics software like VeraChem with the accuracy of quantum mechanical calculations – is particularly impactful for drug discovery. Researchers can rapidly generate numerous molecular conformations, then selectively refine only the most promising candidates using computationally intensive quantum methods. Automating this process on AWS promises to significantly reduce both time and cost. Furthermore, ODU’s work isn’t stopping at workflow optimization; they are actively exploring the integration of artificial intelligence. This project serves as a blueprint for modernizing legacy HPC applications, demonstrating that even complex software can be successfully transitioned to a cloud-native environment with the right support.
