AI Infrastructure

The private cloud and high-performance computing resources of the Department of Informatics and Telematics of Harokopio University of Athens, for education, research and services.

High-Performance Computing cluster

7
compute nodes (hpc01–hpc07)
448
CPU cores
~4.3 TiB
RAM
20+
GPUs (8x H200, 4x H100, 5x A100 and others)

Since 2025, a shared, SLURM-managed High-Performance Computing cluster is also available. It consists of 7 compute nodes (hpc01–hpc07) with 448 CPU cores, ~4.3 TiB of RAM and over 20 GPUs, including 8x NVIDIA H200, 4x NVIDIA H100 and 5x A100 accelerators for AI workloads. It serves research and teaching in machine learning, data science and computational sciences across the university, through shared partitions with preemption-based scheduling.

Status of the HPC cluster partitions and nodes through SLURM (sinfo command)
Server racks of the HPC cluster
Network cabling in the HPC cluster

HERMES LLM

Beyond batch computing, the cluster also hosts HERMES, an on-premise open-weight LLM service (vLLM + LiteLLM, OpenAI-compatible API), giving the university direct, sovereign access to modern AI capabilities on its own hardware.

Status page of the HPC cluster's local LLM service
HERMES running in the command line

Private cloud infrastructure

24
active physical nodes in 4 Proxmox clusters
1,000+
CPU cores
~4.6 TB
RAM
210+ TB
storage

The Department of Informatics and Telematics of Harokopio University of Athens operates an extensive private cloud infrastructure, based on the Proxmox Virtual Environment platform, which supports the hosting and management of virtual machines and services for the department's educational, research and administrative needs. It consists of four Proxmox clusters with 24 active physical nodes, over 1,000 CPU cores, approximately 4.6 TB of RAM and more than 210 TB of storage, and hosts dozens of active virtual machines used for service delivery, application development and testing, and the support of research projects and laboratory activities. The infrastructure provides centralized resource management and flexible, dynamic allocation of CPU, memory and storage in a single, extensible cloud environment, without exclusive dependence on external cloud providers.

Server racks of the department's private cloud
Blade servers of the department's private cloud