Linux Containers for Fun and Profit in HPC
Reid Priedhorsky and Tim Randles
This article outlines options for user-defined software stacks from an HPC perspective. We argue that a lightweight approach based on Linux containers is most suitable for HPC centers because it provides the best balance between maximizing service of user needs and minimizing risks. We discuss how containers work and several implementations, including Charliecloud, our own open-source solution developed at Los Alamos.