Symposium on High Performance Computing (HPC)

Symposium on High Performance Computing (HPC)

on the occasion of the inauguration of Marvin

Research and Technology Center for Detector Physics (FTD, October 20, 2023)

Marvin was put into operation, a big tier 3 HPC cluster and highly important piece of infrastructure for the University of Bonn that many scientists have been waiting for! An ideal occasion to bring together scientists from different fields, whether HPC experts or intensive users, to provide insights into their work to a broad scientific audience. 

Report

Many scientists at the University of Bonn have been eagerly awaiting the launch of the new mainframe computer “Marvin”, which will significantly improve the local HPC capacities for all scientific disciplines.On October 20, 2023, the wait was over and TRA Modelling used the occasion of the official VIP inauguration ceremony of Marvin to extend a complementary invitation for a wide audience - from PhD students to experienced researchers.

Time for discussion and networking: Visitors actively participated in the discussions after each presentation. Between sessions there were longer breaks with coffee, cake, drinks and snacks, providing time and opportunity to talk and discuss with each other. This allowed everyone to exchange ideas with colleagues, including those from different areas of expertise, or simply get to know each other better.

With the help of the broad scientific network of TRA Modelling, seven highly suitable speakers, experts and HPC users alike, could be won to give a talk. Prof. Estela Suarez and Prof. Thomas Luu (both Forschungszentrum Jülich), together with local speakers Prof. Peter Krawitz (University Hospital Bonn), Prof. Julian Klaus (Dept. Earth Sciences), Prof. Carsten Burstedde (Dept. Mathematics), Prof. Petra Mutzel (Dept. Computer Science), as well as Jan Steiner (University IT and Data Center, HRZ) were able to demonstrate the various practical applications of HPC to multiple scientific questions. By covering the core topics of TRA Modelling, mathematics and computer science, and application fields life sciences & medicine, geosciences and physics, the symposium was very much in line with the TRA’s mission to bridge the boundaries between disciplines. With ample time for discussion and networking, 60 participants took the opportunity to meet and gain insights into other research fields.

What Everyone Should Know about Marvin

Jan Steiner gave an very helpful overview of the responsibilities of cluster users and some tips to use Marvin efficiently and effectively, from the perspective of the Marvin support team. Slides can be downloaded here. Latest news about Marvin and HPC in general can be found at the cental hub for High Performance Computing (HPC) at the University of Bonn. 

Review of the program from October 20, 2023


Overview

Session I - Opening by chair Prof. Dr. Jürgen Gall

  • Prof. Dr. Estela Suarez

Computer Science department, University of Bonn, and Jülich Supercomputing Centre, Forschungszentrum Jülich GmbH

  • Prof. Dr. Carsten Burstedde

Institute for Numerical Simulation, University of Bonn

  • Prof. Dr. Thomas Luu

Institute for Advanced Simulation 4, Forschungszentrum Jülich GmbH, and HISKP, University of Bonn

Session II

  • Prof. Dr. Peter Krawitz

Institute for Genomic Statistics and Bioinformatics (IGSB), University Hospital Bonn

  • Prof. Dr. Julian Klaus

Department of Geography, University of Bonn

Session III

  • Prof. Dr. Petra Mutzel

Institute for Computer Science and HPC/A-Lab, University of Bonn

  • Jan Steiner

HRZ HPC Team, University of Bonn

Symposium_High_Perfomance_Computing
© Christina Schiegl

Speakers 

(listed in alphabetic order, program will be hosted by Prof. Dr. Jürgen Gall)

Burstedde_Carsten.jpg
© Carsten Burstedde

Prof. Dr. Carsten Burstedde

Scalable Algorithms (not only) for Geophysics Simulations

Institute for Numerical Simulation (INS)

Member of TRA Modelling

Abstract

not available (Prof. Burstedde was thankfully able to replace a planned lecture of Prof. Dölz, which had to be canceled on a short notice).  


Prof. Dr. Julian Klaus

Predicting water cycle processes under global change

Department of Geography, University of Bonn 

Member of TRA Sustainable Futures

Abstract

Today we are experiencing a global water crisis with more than two billion people lacking access to safe drinking water. The observed non-stationarity of hydrological systems, together with increasing anthropogenic pressure on catchments, are challenging our current capabilities to understand changes in the global water cycle, impacting our capability to predict floods and droughts around the world, and eventually impacting adaption and resilience of society to global change. In this presentation, I will focus on computational approaches to understand and predict processes in the water cycle from small to large scales.

Julian Klaus.jpg
© Julian Klaus

KrawitzPeter.png
© Peter Krawitz

Prof. Dr. Peter Krawitz

Why does Clinical Bioinformatics need Marvin?

Institute for Genomic Statistics and Bioinformatics (IGSB), University Hospital Bonn

Member of TRA Modelling

Abstract

The amount of genetic sequencing data in research and healthcare is growing faster than processing power and storage space. This means that intelligent algorithms for sequence assignment, assembly, and variant detection are crucial. Once these first steps of clinical bioinformatics are done, the data must also be interpreted in the context of the patient's phenotype. Again, an increasing amount of medical images need to be analyzed. To solve these challenges, we need GPU clusters and experts who can handle them.


Prof. Dr. Thomas Luu

HPC and Physics:  Strange Bedfellows or the Perfect Couple? 

Institute for Advanced Simulation 4, Forschungszentrum Jülich GmbH & HISKP, University of Bonn

Abstract

I discuss the general role of HPC in physics, emphasizing how HPC has become an essential tool in performing non-perturbative, strongly-correlated physical systems.  I highlight certain successful examples and discuss various physical problems that can (hopefully) be addressed with future HPC resources.  Finally, I point out how physics in turn benefits HPC, thus providing a means for "closing the loop" and cementing the marriage of HPC and Physics.

Thomas_Luu.png
© Thomas Luu

Mutzel_Petra.jpg
© Petra Mutzel

Prof. Dr. Petra Mutzel

I. Quantum Annealing versus Digital Computing & II. News from the HPC/A-Lab

Institute for Computer Science and HPC/A-Lab, University of Bonn

Member of TRA Modelling

Abstract

Quantum annealing is getting increasing attention in combinatorial optimization. The quantum processing unit of the D-Wave 2000Q is constructed to approximately solve Ising models on so-called Chimera graphs. Ising models are equivalent to quadratic unconstrained binary optimization (QUBO) problems and to maximum cut problems on the associated graphs. 

A conventional digital computer front end is used to transform a given combinatorial optimization problem to one ore more instances of the Ising problem that are fed to the annealing hardware for approximate solutions. These solutions are finally transformed to a solution of the original problem. The transformation steps are problem dependent, and in general quite involved. As an example, we will point out the problems in solving combinatorial optimization problems such as the Traveling Salesman Problem by quantum annealers. 

In the pursuit of a fair comparison, we neglect the digital front end phases and concentrate only on the solution of the Ising instances given to the annealing unit. It has been a matter of discussion in the literature how well the D-Wave hardware performs at its native task. In our experiments we examine how reliably the D-Wave computer can deliver true optimum solutions of the Ising problem. We also compare the annealer's performance in terms of solution time and solution quality with the performance of a heuristic by Alex Selby designed for digital computers and present some surprising results. (This is joint work with Michael Jünger, Elisabeth Lobe, Gerhard Reinelt, Franz Rendl, Giovanni Rinaldi, and Tobias Stollenwerk).

At the end I will briefly present some news from our HPC/A-Lab.


Jan Steiner

What Everyone Should Know about Marvin

HRZ HPC Team

Abstract

An overview of your responsibilities as a cluster user and some tips to use Marvin efficiently and effectively, from the perspective of the Marvin support team.

Jan Steiner
© Jan Steiner

Suarez.jpg
© Estela Suarez

Prof. Dr. Estela Suarez

Modular Supercomputing: heterogeneous hardware for diverse applications

Computer Science department, University of Bonn, and Jülich Supercomputing Centre, Forschungszentrum Jülich GmbH

Member of TRA Modelling

Abstract

The Modular Supercomputing Architecture combines various hardware components like CPUs, GPUs, accelerators, and emerging technologies into specialized compute modules. These modules are linked through a fast network and share a common software stack, creating a versatile system where users can adjust hardware resources by selecting the appropriate nodes per module. This setup enables efficient execution of complex multi-physics and multi-scale simulations across modules and allows users to tailor their hardware resources by selecting the appropriate nodes per module. MSA optimizes system workload distribution by assigning tasks to the most suitable hardware, following each application part's inherent parallelism.


More

High Performance Computing at the University of Bonn

The high performance computing services (HPC) provided by the university of Bonn are an essential service for the campus community. We offer multiple HPC systems, as well as support, consulting and training concerning the use of these systems and scientific programming in general. Our systems Bender, Bonna and, Marvin can be used by all researchers, including students, at Uni Bonn.

Find out more... 

Research and Technology Center for Detector Physics (FTD)

At the FTD, scientists from the Institute of Physics (PI) and the Helmholtz Institute for Radiation and Nuclear Physics (HISKP) of the University of Bonn work on the latest detector technologies for the detection of radiation and particles.

Institute for Numerical Simulation (INS)

The INS is devoted to the design and implementation of computational tools for scientific applications in the Natural and Engineering Sciences, Geosciences and Geodesy, Medicine and Life Sciences, and Economy and Finance.

The INS is the place where Marvin is hosted and extensive reconstruction work had to take place. All future users of Marvin are grateful for this commitment!

HPC: Inauguration of Marvin

The day has finally come: October 20, 2023 the ceremonial inauguration of the new high-performance computer “Marvin” has taken place.

With a symbolic push of a button the rector of the university, Prof. Dr. Dr. hc Michael Hoch, the Vice Rector for Digitalization and Information Management, Prof. Maren Bennewitz, and the Chancellor of the University of Bonn, Mr. Holger Gotschalk, the official starting signal for the  new supercomputer was given at the University of Bonn.

Wird geladen