Optimizing Build Time of Shoc Workloads

Abstract

Running HPC workloads can be challenging due to infrastructure complexity, the learning curve and the specifics of various workload management and scheduling solutions. The paper aims to describe the recent advancements in the Shoc [1] architecture designed for submitting serverless HPC workloads on cloud infrastructure. In particular, the paper discusses the set of architectural advancements in the performance of the build process.