Introduction to EuroCC 3 & the training events

• Nikolaos Bakas (GRNET)

Introduction to High Performance Computing

• Nikolaos Bakas (GRNET)

This talk introduces High Performance Computing (HPC) as a foundation for modern AI and scientific workloads. It explains how clusters combine CPUs, GPUs, nodes, memory, networking, storage, and job schedulers to run demanding computations efficiently. The session covers parallel computing concepts such as cores, threads, tasks, data and task parallelism, then moves into scaling, speedup, efficiency, strong and weak scaling, Amdahl’s Law, and Gustafson’s Law. It also presents key HPC programming models, including OpenMP, MPI, GPUs, and CUDA, before highlighting leading European supercomputers and EuroHPC access routes for researchers and industry. Practical examples help link theory to real workloads.

HPC at the Aristotle University of Thessaloniki

• Paschalis Korosoglou (AUTH)

This opening session welcomes participants to the HPC training program and sets the stage for the sessions that follow. We outline the structure and goals of the program, introduce the topics covered across the individual talks and tutorials, and provide guidance on how to get the most out of the training. Participants get a brief overview of what High Performance Computing is, why it matters for modern research and industry, and what skills they will develop by the end of the program. We also introduce the Aristotle HPC cluster at AUTH serving more than 12 research departments and supporting over 100 scientific publications since 2018, as the primary platform on which the hands-on exercises will be carried out. The training overall requires no prior HPC experience and is suitable for all participants regardless of background.

Practical Slurm: Batch Job Scheduling on HPC clusters

• Vasilis Asteriou (AUTH)

This talk introduces Slurm, a scheduling and workload management system for HPC clusters. It explains how Slurm allocates shared resources among multiple users, manages job queuing, and enables efficient parallel job execution. The session covers Slurm's multifactor priority scheduling and backfill algorithms that optimize resource utilization. Practical examples demonstrate job submission for various workloads including serial jobs, multi-core CPU tasks, MPI parallelization (using Gromacs), memory-intensive applications, GPU computing, and licensed software. Key Slurm commands for monitoring and managing jobs are presented, providing researchers with essential tools for accessing and optimizing HPC resources on the Aristotle cluster.

Open OnDemand: the HPC Web Portal

• Alexandra Charalambidou (AUTH)
• Anastasia Makri (AUTH)

Open OnDemand is a web portal application that enables HPC access through any modern browser, eliminating the need for SSH clients or command-line interfaces. The presentation covers the platform's core capabilities: file management, interactive applications like Jupyter and RStudio directly on compute nodes, and a web-based shell. The feature demo showcases virtual desktops, VS Code, MATLAB, and customized user experiences for different user types. The Project Manager, a recent feature, enables researchers to create and organize research projects while building, submitting, and monitoring batch jobs through an intuitive interface.

XDMoD Usage Reports

• Alexandra Charalambidou (AUTH)

XDMoD (XD Metrics on Demand) is an open-source web portal that provides comprehensive monitoring and analysis of High Performance Computing (HPC) resource usage.