Forschungsthemen
[MA] Design and Implementation of a CI/CD Pipeline for High Performance Computing Architectures
High Performance Computing (HPC) architectures are increasingly utilized for a variety of scientific and engineering applications such as Machine Learning (ML) tasks that require large-scale computing resources. However, deploying and managing software on HPC clusters is a challenging task due to the heterogeneity of the software and hardware environments. Nowadays, Continuous Integration, Delivery and Deployment (CI/CD) is widely used to deploy rapidly evolving software solutions quickly and easily. This thesis aims to design a general-purpose CI/CD pipeline for HPC architectures that enables quick and easy deployment of different ML software stacks on a set of HPC resources using a standardized and automated process. The approach will be implemented as a prototype and evaluated based on given use cases. This will enable scientists and developers to easily package their software artifacts and deploy on a set of HPC resources, improving the software qualities such as reliability and reproducibility of the software deployment process.
Betreuer: Karsten Wendt