Refactoring a statistical package for demanding memory loads: Adapting R for high performance telemetry data analytics

An R library application employing distinct resampling and multithreaded overlap routines demanded memory beyond the limits of a desktop workstation. Initial efforts revealed that the necessary memory allocations for the dataset output from the resampling exceeded that available on a high memory node, making code refactoring necessary. Subsequent in-depth analysis of the employed package revealed that by combining and multithreading bootstrap sampling and overlap calculations, the memory footprint was reduced to a manageable size. However, the computation time was substantially increased. Further development, decomposing the resampling step into two routines, minimized the impact of a rate-limiting step, restoring shorter runtimes while maintaining the smaller memory footprint.