The rapid capabilities of cloud computing allow for complex computational chemistry tasks to be completed in days instead of months.
Some computational challenges are so substantial that a comprehensive approach is required. This is the strategy employed by a varied group of scientists and computing specialists from the Department of Energy’s Pacific Northwest National Laboratory, alongside partners from Microsoft as well as other national laboratories and academic institutions, aimed at making new cloud computing resources more accessible.
The initiative, discussed in a recent peer-reviewed article, offers a plan to transition scientific computing resources into a sustainable framework that grows alongside advancements in technology. The research team showcased that cloud computing acts as a responsive and flexible supplement to traditional high-performance computing centers that have been crucial for scientific research for many years.
“This represents a completely new model for scientific computing,” stated PNNL computational chemist Karol Kowalski, who spearheaded the interdisciplinary initiative. “We have illustrated the feasibility of integrating software as a service with cloud computing resources. This initial proof-of-concept indicates that cloud computing can offer various options to enhance and support high-performance computing in addressing intricate scientific issues.”
Cloud-based Sustainable Software
The cloud has evolved significantly beyond being merely a storage solution for photos and documents. The computing sector is transitioning to offering computation as a service to various industries, including finance and pharmaceuticals. In this project, the team concentrated on migrating resource-intensive algorithms used for assessing the viability of new industrial chemicals, advanced polymers, surface coatings, and many other uses to the cloud.
This initiative, named Transferring Exascale Computational Chemistry to Cloud Computing Environment and Emerging Hardware Technologies (TEC4), builds upon the growing demand within the computational chemistry sector to deliver computational resources to users, acknowledging the ongoing need for software adjustments to satisfy both scientific demands and evolving hardware.
In their latest perspective piece, the team shares insights and performance data regarding both legacy computing algorithms, such as the well-known NWChem software originally developed at PNNL, and the latest applications designed to leverage advanced graphics processing unit (GPU) architectures. Their findings revealed that the speed and flexibility offered by cloud computing enable advanced computational chemistry tasks to be completed in days instead of months.
“Microsoft aims to empower the scientific community to expedite scientific discovery,” expressed Nathan Baker, product leader for Microsoft’s Azure Quantum Elements. “This partnership with PNNL exemplifies how modern AI [artificial intelligence] and HPC tools can enhance computational chemistry.”
Addressing Critical Energy Solutions
In the last decade, computational chemistry has demonstrated its capacity to tackle intricate scientific problems while also guiding and interpreting experiments and facilitating predictions. The most challenging issues are best resolved using the resources offered at the DOE’s leading computing facilities, particularly with exascale computing capabilities.
As tools and techniques have progressed, so have the costs and duration associated with reaching solutions. The leadership team at TEC4 recognized that the combination of cloud computing and industry partnerships provides an avenue for a broader range of problem-solving opportunities.
For instance, the team utilized Microsoft Azure and advanced workflows to explore the molecular dynamics involved in intricate chemistry problems. These simulations are invaluable for investigating complex reactions that are difficult to observe through experiments. This robust tool, employed to analyze atomic-level molecular interactions, demands extensive computational resources due to its intricacy. The research team demonstrated a method for breaking down the persistent environmental contaminant perfluorooctanoic acid, showcasing how computational chemistry can inform practical strategies in environmental cleanup.
“We envision a system of use cases ranging from basic to advanced tasks that utilize GPU-based computing, which is currently being widely adopted for AI and machine learning applications,” Kowalski explained. “We aim to allow users to leverage different computing tiers, charging only for what is necessary and integrating software with compute access. This marks the initial step towards that envisioned future.”
A Cloud Computing Ecosystem
The team is actively seeking new collaborators from both development and user communities to establish a testing base for the new cloud ecosystem.
“We are creating a portfolio of codes,” Kowalski added. “The objective is to build a community around this initiative.”
To support this goal, the team plans to train a group of students proficient in these tools, addressing the need for scientists who can advance computational techniques. This collaboration has led to the development of a new course at the University of Texas at El Paso, with contributions from Central Michigan University and PNNL set to begin in fall 2024.
