Abstract text (incl. figure legends and references)
The constant development of faster electron detectors and microscope software and the analysis of flexible and dynamic macromolecular complexes demand powerful computing resources. While the microscope operation and image processing are getting increasingly automated and require less manual input, the design and deployment of an HPC system, suitable to fulfill such demands, is still very tedious and can quickly become overwhelming, especially for not professional system administrators.
Here we describe the design choices for our recent cryo-EM HPC system and explain how we automated its installation and deployment with a combination of Metal-As-A-Service (MAAS) and Ansible. The cluster incorporates multi-GPU nodes, highspeed Interconnection (InfiniBand HDR) and a parallel file system (BeeGFS). Our approach can be easily adapted to different storage and hardware systems and enables inexperienced personnel to set up an HPC with little time.