SimplifyHPC: A Framework to Simplify HPC Codes Deployment

EP Analytics will develop and commercialize a general-purpose and cross-platform solution that streamlines the deployment of complex modeling and simulation (Mamp;S)nbsp;applications on a diverse set of computing resources (local desktop/laptops, all major cloud-service providers and traditional on-premise HPC clusters). The proposed solution, SimplifyHPC,nbsp;is a desktop application that brings together all the tools necessary for Mamp;S scientist andnbsp;will consist of anbsp;built-in performance monitoring ensuring optimal deployment of applications.NASA scientists and engineers increasingly rely on large-scale numerical simulations to advance understanding of complex Earth and astrophysical systems and to conduct High-Fidelity aerospace engineering analyses. Building and running most numerical simulation codes, however, are challenging tasks that involve resolving multiple dependencies (e.g., on external frameworks such as PETSc and NETCDF), and selecting optimal runtime parameters (e.g., number of OpenMP threads per MPI rank). Even for expert HPC users, these deployment steps pose significant barriers to executing simulation codes efficiently.nbsp;EP Analytics proposes to develop an easy-to-use shrink-wrap solution, SimplifyHPC, that is designed to address the aforementioned complexities and remove the barriers to entry for prospective users within NASA and beyond. SimplifyHPC will deploy methodologies to collect lightweight performance reports, which can help uncover needed cost-saving optimizations and ensure efficient execution of HPC applications, thereby adding to the bottom line of SimplifyHPC customers.nbsp;nbsp;