Organization: Los Alamos National Laboratory
Year: 2018

or software environments—to high performance computing (HPC) resources, granting world-class computational power to programs that otherwise would have been difficult or impossible to deploy. Charliecloud unshackles researchers from traditional software environments, system administrator permissions, and bulky, bugridden programs, and allows them to harness HPC resources for any of their research needs. Like other container runtimes, Charliecloud uses images—a user-defined package that holds the environment needed to run an application— to create containers. Charliecloud’s fundamental innovation is that it separates building and sharing images from running them, allowing building and sharing to occur in a sandbox that is separate from HPC center resources. This can be a user workstation, a virtual machine, or anywhere else that is safe for users to have the privileged access they need to build and share images. In this sandbox, the user runs industry-standard Docker—or any other tool of their choice—to create their desired image. The image is then easily transferred to HPC resources where researchers can run applications using the lightweight, unprivileged Charliecloud runtime, which only knows how to run containers. The Charliecloud runtime is accessible anywhere: within the sandbox for development and testing, on HPC resources, or somewhere else.

For more info: