A fast optimizer for HPC technology insertion

To maximize the program's overall capability, the HPC Modernization Program annually acquires new hardware by carefully benchmarking new viable platforms, gathering corresponding pricing, and optimizing the selection of new machines based on scoring, pricing, and a projected work allocation among computational technology areas. Until recently, optimization had been performed using Microsoft Excel's Solver on a single-processor personal computer with a large time-to-solution on the order of hours for small problems and days for large problems. In response, a fast multiple-processor optimizer specific to the needs of the technology insertion process was constructed. This parallel acquisition optimizer is described in detail, and its results are compared to Solver's for a number of practical test cases.