A heterogeneous terascale computing cluster for the development and efficient implementation of high order numerical methods
Abstract
The development of efficient and robust numerical algorithms suitable for modern high performance computing architectures has become a major ingredient needed for simulation of complex multiscale problems. The continual challenge in high performance scientific computing is the need for cutting-edge equipment to enable the application-driven development of innovative numericalmethods to simulate complex multiscale physical problems, and the related education and training a mathematical and computational workforce. This proposal aims to directly meet these challenges by acquiring a new heterogeneous terascale computing cluster, which will become a shared campus research instrument for an inter- and multidisciplinary group of mathematicians and computational scientists and engineers, and their undergraduate and graduate students. The proposed cluster will contain 73 nodes, with multi-core processors connected over a high-speed network. This instrument will be used to develop hybrid parallel and accelerator-based computing algorithms by incorporating 15 nodes with NVIDIA Tesla GPGPUs or Intel Xeon Phi processors. In addition, a modern storage system with 132TB of usable space will be connected to all the compute nodes. This computing cluster will be used to foster the development and testing of novel mathematical methods suitable for large scale parallel scientific computing and GPU acceleration. These novel mathematical algorithms will be designed forcutting-edge high performance, distributed-memory parallel, and GPU accelerated numerical simulations of ordinary and partial differential equations in a variety of multidimensional, multiscale, and multiphysics scientific applications, and will further the education and training of undergraduate and graduate students with the goal of developing the next generation of experts in scientific computing.Using this computing cluster, we propose the application-driven development of numerical methods suitable for HPC simulations. In particular, four research projects will be enabled by this cluster: Project 1 focuses on the development of optimal time-stepping methods for a variety of complex multiscale physical problems including gut microbiome dynamics, quantum chemistry problems,oceanographic and multi-phase flows, and black hole simulations. Project 2 deals with the development of reduced order modeling approaches in the reduced basis family. The proposed cluster will enable rigorous development and application of reduced basis methods for a variety of applications, including geometric engineering design under uncertainty, multiphase flows, and data assimilation in oceanographic simulations. Project 3 involves the development, analysis, and implementation of superconvergent hybridizable discontinuous Galerkin (HDG) methods, which hold the promise of being highly efficient as well as remarkably scalable, for wave problems that feature multiple scales. Project 4 focuses on the development of the methods in Projects 1-3 for GPGPUs and next generation computing paradigms.The availability of this cluster will allow for the training of the next generation of high performance computing experts, who are currently graduate or undergraduate students, and will support a new interdisciplinary Ph.D. program in computational science and engineering at UMass Dartmouth. Graduate students will benefit from the opportunity to develop cutting-edge parallel and GPUaccelerated algorithms for a variety of challenging physical problems, in a truly multi-disciplinary setting. The PIs will integrate undergraduate students into their research programs, in the effort to prepare undergraduate students for, and engage them in, sustained research experiences in computational mathematics.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 10, 2018
- Source ID
- N000141812255
Entities
People
- Sigal Gottlieb
Organizations
- Office of Naval Research
- United States Navy
- University of Massachusetts