TRANCE: Translating RADICAL Abstractions to Nascent Computing Environments

With the sudden onset of COVID19 and the need for an “immediate” response, Brookhaven and Argonne National Laboratories developed a computational infrastructure to support a high-throughput virtual drug screening pipeline that synergized with DOE experimental facilities. The resulting infrastructure evolved to become the workhorse of the DOE’s National Virtual Biotechnology Laboratory (NVBL) Medical Therapeutics (MT) project and came to be known as Integrated Modeling PipEline for COVID Cure by Assessing Better LEads (IMPECCABLE). IMPECCABLE was used on DOE leadership platforms to screen an un- precedented number of drug-like candidates and was significant in achieving the scientific goals of NVBL’s MT project. IMPECCABLE both used and drove the development of the RADICAL-Cybertools (RCT) to support scale, portability, and workload and platform heterogeneity. The fact that IMPECCABLE was functionally ready in four weeks, and was gradually scaled to meet the scientific and computational requirements in a matter of months is a direct consequence of the design and implementation details of RCT. The performance and portability of RCT on Leadership platforms has been well documented. For ex- ample, RCT was also a central component of the ACM 2020 Gordon Bell COVID-19 AWARD (co-led by Brookhaven) wherein it played a critical role in enabling the high-performance workflows that provided scientific insight leading to the GB COVID19 Award (at SC’20). Several barriers to performance and portability remain: RCT are not usable on Cloud platforms, nor are they positioned to leverage the evolving convergence of HPC and Cloud systems. Reflecting its origins, RCT has grown to emphasize raw performance and narrow usage mode over usability, extensibility and embracing recently ascendent usage modes and programming models. Incomputable Systems LLC proposes to extend and enhance RCT to overcome these natural but fundamental limitations. We plan to leverage the recent successful application to non-traditional HPC scenarios, to repackage RCT into a capability ready to support discovery and innovation in commercial settings and constraints. Incomputable will take the capabilities of RCT and expose them to the forces of the markets and wider technology ecosystem. This project will enable Incomputable LLC to translate the scalable computing principles and practice embodied in ASCR-funded RCT to new computing platforms, and commercial scenarios and services, while developing and hardening a viable commercialization and uptake strategy and plan. Incomputable Systems LLC will make RCT easier to use by a wide variety of industries (in particular the pharmaceutical and other molecular design industry), on a diverse range of platforms (major cloud providers and private clouds), as well as provide the basis for future advanced services (i.e., commercial offerings of RCT). We have the support and commitment of the RADICAL team based at Brookhaven (led by Shantenu Jha, Chair of Computation and Data Driven Discovery and PI of RADICAL-Cybertools project).