Computing System for the Simulation of Multiphases Proceses
Abstract
This DURIP proposal requests funds for a high performance computing system to support RensselaerÕs Scientific Computation Research Center (SCOREC) research on the development and application of simulation technologies to model multiphase processes. This research is supported by the U.S. Army Research Office grant entitled ÒMultiscale Methods for the Reliable Simulation of Multiphase ProcessesÓ, W911NF-14-1-0301, with Professors Assad Oberai, Mark Shephard, Onkar Sahni as Principal Investigators. The program manager is Dr. Joseph D. Myers, Division Chief of the Mathematical Sciences Division of the Army Research Office. The core computational component of the requested system is a 16-node computer cluster build on Intel s new Xeon Phi processor known as Knight s Landing (KNL). The requested cluster is six times faster, in terms of both compute power and data access, than SCORECÕs fastest current cluster. The second system component is a large memory data server that will be used primarily for cases where collaborators provide large serial input sets. The third component is a storage system with twenty four 900GB 10,000 RPM disks with servers and associated virtual machine system to provide high speed access to storage for the KNL cluster and data server. The last component is five visualization workstations to support students and postdocs in the effective development of the simulation tools and to support users in the effective application of the resulting software on problems of interest. The requested research instrument will be fully integrated into SCORECÕs computing infrastructure. The multiphase simulation techniques and associated software are being developed in collaboration with DoD researchers at Benet Labs, under US ArmyÕs Armament Research and Development and Engineering Center (ARDEC), and the Coastal and Hydraulics Laboratory under US ArmyÕs Engineer Research and Development Center (ERDEC). ARDEC researchers are interested is using the technologies to model interior and exterior ballistic processes associated with firearms and large cannons. ERDEC researchers are interested in two-phase flow problems with moving interfaces and structures. Other areas of application of multiphase methods of interest to the DoD include fluid structure interactions for various aerial vehicles, combustion systems, phases change heat transfer devices, material processing for advanced structures and electronics, and manufacturing processes from mold filling, to additive manufacture, to classic subtractive manufacturing. Many of these areas of application are relevant to industry for both military and civilian applications. Even when using the adaptive methods being developed in this project, the accurate simulation of the multiphase problems of interest requires meshes with hundreds of millions, to billions, of unknowns solved over several thousands of time steps. To properly address these applications, all components of the simulation workflow must operate in parallel on the same massively parallel system. With the current rapid evolution of massively parallel computer architectures it is critical our parallel algorithms are developed specifically considering the new architectures to ensure scalability and performance. Such developments are best supported by directly developing on systems based on these new architectures. This is why this proposal is requesting a computing system based on the latest multicore architecture.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 14, 2019
- Source ID
- W911NF1810228
Entities
People
- Mark Shephard
Organizations
- Army Contracting Command
- Rensselaer Polytechnic Institute
- United States Army