Containerization on Petascale HPC Clusters

Containerization technologies provide a mechanism to encapsulate applications and many of their dependencies, facilitating software portability and reproducibility on HPC systems. However, in order to access many of the architectural features that enable HPC system performance, compatibility between certain components of the container and host are required, resulting in a trade-off between portability and performance. In this work, we discuss our early experiences running three state-of-the-art containerization technologies on the petascale Frontera system. We present how we build the containers to ensure performance and security and their performance at scale.We ran microbenchmarks at a scale of 4,096 nodes and demonstrate the near-native performance and minimal memory overheads by the containerized environments at 70,000 processes on 1,296 nodes with a scientific application MILC - a quantum chromodynamics code.