AIME Compute Server
The AIME Compute Server is a high-performance research server of the Chair of Computer Engineering for compute-intensive student and research projects. It provides isolated LXC container environments with controlled resource allocation and optional GPU access.
Purpose & Motivation
The server provides central compute infrastructure for workloads that exceed typical student laptops or office workstations, such as AI experiments, sensor-data processing, computer vision, and drone-related data analysis. It helps students and researchers work in reproducible, isolated environments without setting up dedicated hardware for each project.
Technical Description
The system runs Ubuntu 24.04 LTS and uses LXC/LXD containers to separate student and project environments. Containers can be provisioned with a lightweight script to reduce setup and cleanup effort. Students access their environments via university credentials and can directly use their AFS storage.
Hardware overview:
- Server platform: ASRockRack SIENAD8-2L2T
- CPU: AMD EPYC 8224P, 24 cores / 48 threads
- Memory: 192 GB RAM
- GPU: NVIDIA GeForce RTX 5090 with 32 GB VRAM
- Storage: 2 × Samsung SSD 990 PRO 2 TB
- Container backend: LXD 5.21 LTS with ZFS storage
Capabilities
-
Isolated Linux containers for students and projects
-
Controlled allocation of CPU, memory, storage, and optional GPU access
-
Shared GPU usage for AI and computer-vision workloads, coordinated between users
-
URZ-based authentication
-
Direct access to AFS storage
-
Fast provisioning and removal of student containers through a simplified script-based workflow
Research Context
The server supports research and teaching activities related to AI, sensor-data processing, computer vision, and drone applications. It is especially relevant for student theses and project work that require reproducible compute environments or GPU-supported experiments.
Related research context includes the SmartCargoDrone project and infrastructure around the TUC SmartCityCloud, where sensor and research data can be stored, visualized, and later processed.
