Logo

HPC @ Uni.lu

High Performance Computing in Luxembourg

Due to COVID restrictions, this edition of the UL HPC School will entirely take place remotely.
This tenth edition of the UL HPC School will take place on:
December 15th and 16th, 2020, Remotely (via Webex)


   Registration Portal

The University of Luxembourg operates since 2007 a High Performance Computing platform which currently features a total of 514 nodes for a cumulative computing power estimated at 2,76 PFlops.

The total raw storage capacity shared within the clusters is currently estimated above 10 PB.

The effective usage of this complex platform is not an easy task and requires both talent and system skills to understand the impact of our personal workflow on the global performance of the system.

The UL HPC team, together with leading computational scientists of the UL and HPC technologists will offer instructions, hands-on and guided sessions (including several new ones) on a variety of topics representative of research activities and domains present at the university, including:

  • Access to and interaction with the UL HPC infrastructures
  • Monitoring, Debugging & Profiling
  • HPC workflow management (for sequential and parallel tasks)
  • HPC programming and usage of the main software available on the platform, with dedicated sessions directed towards Scalable Science (OpenMP/MPI), GPU trainings (CUDA programming, Neural Networks), R, Python, Big Data analytics, Deep and Machine learning etc.
  • Virtualisation with Singularity containers on the clusters

The aim is to cover basic as well as advanced usage of the platform. Whether you have no HPC experience or are an advanced user, don’t miss this unique opportunity to learn more about the efficient usage of the system.

In all cases, we encourage you to attend the inaugural keynote (scheduled on Tuesday, December 15th 2020, 09h00), where a general update of the HPC and Big Data developments at the University, national and European levels will be provided/ Another keynote is also scheduled on Wednesday, December 16th 2020 (11h30), addressing HPC challenges tied to storage data management.

Important Pre-registration is required. No on-site registration will be accepted and the number of seats is limited Participation to this edition is free of charge for UL and partnered public research institutes members. A specific edition for external partners will be scheduled afterwards.

List of the practical sessions

This workshop is composed of several keynotes and practical sessions. The below table offers the list of these sessions covered during this event – you will find them online on ulhpc-tutorials.rtfd.io.

PS Description Speaker Duration
1 Preliminaries (SSH - OpenOnDemand) A. Olloh, T. Valette 1h30
2 Getting Started 2.0 SLURM, performance engineering and basic launchers H. Cartiaux 2h00
3 HPC Management of Sequential and Embarrassingly parallel jobs S. Varrette 1h30
4 Prototyping with Python (incl. jupyter notebook), Deep Learning basics (CPU, GPU) F. Pinel, 2h00
5 Scalable Science with OpenMP/MPI E. Krishnasamy 2h00
6 HPC Containers with Singularity E. Kieffer 1h00
7a Introduction to GPU programming with CUDA (Part I) F. Pinel 1h00
7b Introduction to GPU programming with CUDA (Part II) L. Koutsantonis 1h00
8 Big Data Analytics S. Varrette 1h00
9 Multi-GPU Training of Neural Networks C. Hundt (Nvidia) 1h30
10 R - statistical computing A. Ginolhac 1h30

Program

All sessions will take place on Webex – invitation links will be provided to registered participants.

All tutorials proposed as practical sessions will be available on GitHub.

Practical information

Due to the COVID restrictions, the ULHPC School will be performed REMOTELY through WebEx sessions.

Access to the sessions will be provided to registered participants, together with a link to access a dedicated Slack workspace we will use for live support suring the session.