CreativeC has provided a number of custom built single node, and small cluster systems for specialized simulation at LANSCE. These systems incorporate the latest technologies in processing, memory, networking, and data storage. Since these technologies are not mainstream, or incorporated in mass production yet, building these systems for large scale manufacturers is actually more difficult and expensive than for a small local business. Also it is the iterative bench-marking of actual workloads which leads to a custom best blend of technologies specific to a certain workload.
This is the concept of the Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) Instrument project funded by Department of Energy (DOE) Office of Science, and New Mexico Small Business Assistance (NM SBA) grants. The concept has also proven for a number of projects at LANSCE. CreativeC configured and optimized several such systems for the facility.
Frank L Krawczyk, LANL, AOT-HPE
Our project uses simulations to understand deformation and failure of materials when they are subjected to a shock wave. The simulations are performed on tens of millions of atoms using molecular-dynamics. Based on time to solution, and cost of equipment, our LAMMPS Instrument is about 8X more efficient than current LANL computing, for a 20M atom shock simulation.
Dr. Saryu Fensin, LANL, MST-8