Impact of Functional Package Manager Over a Non-Traditional High Performance Computing System

Nowadays, practitioners in both industrial and academic circles are using embedded systems that interact with each other to tackle high performance computing workloads, but with individually less capable devices. Smart cities, autonomous machines and agriculture applications are some examples of these systems working in different fields. These systems run several applications that we want to handle in a simple way. There are several approaches for application management and deployment: containers, virtualization and package manager are some examples. In order to monitor the performance of these systems, we evaluated the impact of the package manager method in an embedded systems test bench. We presented an experiment using an 8 NVIDIA Jetson TX1 cluster with Nix package manager and Jacobi Benchmark, showing that this method has a negligible impact performance on execution time against its native version.