hpc.bw – Competence Platform for Software Efficiency and Supercomputing

It’s a pleasure for us to announce that we have a new HPCCP website! Please feel invited to visit our new website:

https://www.hsu-hh.de/hpccp/

Screenshot: HPCCP-Webseite
HPCCP-Webseite (hpc.bw / dtec.bw)

The project „hpc.bw – Competence Platform for Software Efficiency and Supercomputing“ has started in April 2021 and will run for four years until December 2024. It is funded within the framework of dtec.bw and is located in the KoDiA research cluster. The aim of hpc.bw is to strengthen innovative cross-location research in the field of high performance computing and to promote the transfer of relevant expertise to a wide range of disciplines.

Project leader:

Prof. Dr. Philipp Neumann, High Performance Computing, Helmut Schmidt University

Multiplicator team:

Prof. Dr. Sabine Schmidt-Lauff, Weiterbildung und Lebenslanges Lernen, Helmut Schmidt University

Prof. Dr. Thomas Carraro, Angewandte Mathematik, Helmut Schmidt University

Prof. Dr. Marcus Stiemer, Theoretische Feldtechnik und Numerische Feldberechnung, Helmut Schmidt University

Prof. Dr. Andreas Fink, BWL, insbesondere Wirtschaftsinformatik, Helmut Schmidt University

Prof. Dr. Alexander Popp, Computer-Based Simulation, Universität der Bundeswehr München


Do you have any questions about the project? You can reach the project team at [email protected] .

Newsletter hpc.bw

Welcome to the newsletter of the dtec.bw project hpc.bw.

If you want to subscribe to the newsletter, please send a message with subject line “Subscription hpc.bw Newsletter” to [email protected].

Newsletter HPC 23/02

Past Newsletter: 

Newsletter hpc.bw 01/2023
Newsletter hpc.bw 04/2022
Newsletter hpc.bw 03/2022
Newsletter hpc.bw 02/2022

Newsletter hpc.be 01/2022

hpc.bw – Competence Platform for Software Efficiency and High Performance Computing

Link to the project homepage: https://dtecbw.de/home/forschung/hsu/projekt-hpcbw/projekt-hpcbw

High Performance Computing (HPC) deals with the development of efficient software, taking into account algorithms, the use of appropriate programming languages and programming environments, as well as the code-specific optimization of computing instructions. As a result of advancing digitalization, HPC now represents an essential building block of science, research, and development, which many disciplines — both within and outside the armed forces — depend upon. Combinatorial optimization problems in logistics and production, numerical simulations in engineering or image analysis, and machine learning methods with real-time requirements in medical diagnostics are just three examples with high HPC relevance.

However, the potential of existing software solutions is often not fully exploited because computing resources are not always used efficiently. There could be various reasons for this, such as the time required for corresponding software optimizations, or the lack of HPC expertise. Ergo, there are gaps between the actual availability of computing power, the knowledge of computing potentials and the usability of the underlying systems. This results in reduced productivity when using digital research methods.

The aim of the hpc.bw project is to strengthen innovative cross-site HPC research at the universities of the armed forces and to accelerate the transfer of HPC knowledge to a wide range of disciplines in order to

·         sustainably strengthen research and development in the respective disciplines,

·         to promote interdisciplinary exchange between HPC-related problems,

·         derive new research questions for HPC from the various discipline-specific problems and answer them, and

·         to establish a joint HPC competence platform for users within and outside the armed forces.

To this end, a program will be set up, which will focus on disseminating information on HPC and offering contact points to HPC experts at the armed forces universities. This will be supplemented by further training on the use of HPC resources. In order to increase (research) productivity, software packages used by hpc.bw personnel are to be made more efficient and prepared for use on HPC platforms such as supercomputers.

These activities are supported by the establishment of a container-based HPC computing center. In order to ensure sustainability in HPC hardware and software, continuous evaluations of innovative HPC systems (in the context of benchmarks) as well as the support of HPC software projects with regard to software design, testing, etc., are planned.

Entries of the project in directories and specialized databases:

Seminar series „Computation & Data“

We establish a novel seminar series “Computation & Data” at HSU. The goal of this interdisciplinary seminar is to bring together researchers and foster exchange on the development of algorithms, methods and software with regard to (but not necessarily limited to):
• Scientific Computing & Computational Methods: Numerical methods and optimization, code coupling, etc., and their application to challenging discipline-specific problems
• Data science: machine learning, data analysis, statistical methods, etc., and their application to challenging discipline-specific problems
• Computational infrastructure & Hardware-aware programming: code optimization and parallelization, reconfigurable hardware, distributed/cloud/edge computing, etc.
The seminar series is scheduled for the last Thursday every month, 14:30-16:00, with 1-2 presentations, each 25 minutes, per session.

Overview of the lectures WT 2023

Seminar sessions WT2023

If you are interested in joining, please send an e-mail to [email protected] with the subject line „Subscription Seminar Computation & Data“.

Next Seminar: 23.02.2023 14.30 – 16.00 HSU H1 room 308 and via MS Teams

Link to seminar https://teams.microsoft.com/l/meetup-join/19:[email protected]/1668524366704?context=%7B%22Tid%22:%225832f73f-b0fa-45a0-80d9-7e32bd7fa822%22,%22Oid%22:%226003e630-5d1e-4ccb-a8ab-e5d81b1bf8b1%22%7D

Presentation I: Hybrid Finite Element/Deep Neural Networks Methods for Accelerating Fluid-Dynamics Simulations

Nils Margenberg (HSU)

Presentation II: New developments within the Macro/Micro Coupling software MaMiCo for highly parallelized multiscale flow simulation

Louis Viot (HSU)

2nd Call for Projects for Performance Engineering

We cordially invite ALL departments at the Universities of the Armed Forces to apply with HPC-relevant
research proposals. In addition to outlines of individual research interests, there is also the possibility of
submitting proposals for planned or existing research projects.

Please fill in the application form and submit it by 14.02.2023 via e-mail to [email protected].
Please direct any queries to the same address or to Prof. Philipp Neumann, Tel. 040/6541-2723.

We look forward to your contributions!

Past call

2nd Call for projects for performance engineering
Call for projects for performance engineering

The following projects are supported by hpc.bw:

  • The 2-stage no-wait hybrid flow shop scheduling problem (Wiedra, HSU)
  • Single machine scheduling with position dependent maintenance (Wiedra, HSU)
  • Case Study “Personnel Scheduling in RoRo Terminals within the project Digitalization and Technology Research Center of the Bundeswehr/German Federal Armed Forces (dtec.bw) (Hipp, HSU)
  • Optimization of an IGA Code in MATLAB for parallel computing (Dr.Ing. Georgios Michaloudis, UniBw M)
  • Monte Carlo simulations of real fluids (Univ.-Prof. Dr.Ing. Karsten Meier, HSU)
  • Enabling High-Throughput Studies of Reactive Materials (Christopher Lange, HSU)
  • benEFIT- Numerical simulation of non-destructive testing in concrete (Fabian Dethof, HSU)

Case Study “Personnel Scheduling in RoRo Terminals within the project Digitalization and Technology Research Center of the Bundeswehr/German Federal Armed Forces (dtec.bw)

M. Sc. Andreas Hipp (Chair of Business Administration, HSU)

In Roll-on/Roll-off (RoRo) terminals, maritime transshipment of vehicles is carried out as pre- and post-processing for maritime transportation with RoRo ships. In the context of this case study, we focus on personnel scheduling in RoRo terminals, in particular scheduling of mono- and multi-skilled personnel who drive vehicles from parking areas of a RoRo terminal onto decks of a RoRo ship, lash vehicles for security during maritime transportation and drive shuttles for moving personnel between decks and parking areas. Groups of vehicles, which are stored in these parking areas and which are planned to be loaded, are called batches. Because unloadings mirror loadings in reverse order, we focus on loadings for simplicity. Motivated by sudden incidents concerning the personnel, such as mass COVID-19 disease or strikes, we consider the specific case that internal personnel are insufficient for loading all batches as planned, even with expansion by external personnel from temporary employment agencies. For quick response decision making, we use weightings to decide whether batches will be loaded or not. We formulate a Constraint Programming (CP) model with the hierarchical objectives to maximize the sum of weightings of loaded batches and to minimize the number of used external personnel. We solve generated datasets of RoRo terminals with the CP Optimizer and propose a model-based matheuristic, which is inspired by a rule-based heuristic mirroring the human planner’s manual personnel scheduling.

Together with the project group hpc.bw we want to use the supercomputer HSUper to find optimal solutions for the generated datasets. So far, the exact algorithm of the CP Optimizer in combination with commercial hardware could not provide such solutions in acceptable runtimes. Furthermore, we want to analyze the optimal solutions and compare them with those of the proposed model-based matheuristic.

The 2-stage no-wait hybrid flow shop scheduling problem

Frank Wiedra, M.Sc. (Chair for Business Administration, HSU)

ybrid flow shop (HFS) scheduling problems as a flow shop production layout adding parallel machines on at least one stage of operation represent popular systems in real-world production environments.  In our work, we focus on the no-wait HFS scheduling problem. The no-wait constraint leads to the requirement of processing jobs continuously without any waiting time between subsequent stages as soon as they enter the production system. We focus on a two-stage HFS with two machines in the first stage and a single machine in the second stage, as found in steel or chemical industries. It is proved that even the 2-stage HFS is NP-hard so that fast, high-quality solutions may not be readily generated in practical cases. We apply a decomposition approach using several typical heuristics to find a suitable job sequence for our layout. Initially, assignment rules were implemented to schedule the jobs on the machines in the determined sequence – as it is normally done for this type of problem. This 2-stage solution approach is applied to a testbed with up to 200 jobs.

Although the tackled HFS scheduling problem is NP-hard, we found out that the assignment problem with a given job sequence can be solved to optimality for the 2-stage no-wait HFS layout. Together with the hpc.bw project group at the Helmut Schmidt University, we want to generate efficient code for our formulated assignment algorithm and apply it to the entire testbed. Finally, we aim to gain insight into the speed and performance gaps between the exact algorithm and the heuristic assignment rules by running the instances on the HPC servers.

Single machine scheduling with position dependent maintenance

Frank Wiedra, M.Sc. (Chair for Business Administration, HSU)

Machine wear and tear is a typical problem that must be dealt with in real-world production environments. In scheduling research since the 1980s, time-dependent deterioration is usually identified as the main factor for machine wear. In our work, we focus on machine deterioration that is not primarily influenced by time, but by the number of jobs. Each job, regardless of its processing time, leads to the same deterioration, which is tantamount to position-dependent (pd) maintenance. After a given number of jobs, the machine is completely worn out and must be restored by performing a maintenance operation of fixed length. This maintenance operation can also be performed earlier, which leads to the same effect for the machine – it is completely restored. In general, pd maintenance is relevant to any real-world setting where the number of jobs causes the main deterioration and not their length, such as the wear and tear of batteries in automobiles with combustion engines or the landing gear of aircrafts. So far, only a few studies have addressed scheduling problems with pd maintenance yet, see, e.g., Drozdowski et al. (2017), who introduce this type of deterioration.

In this project, we deal with the single machine scheduling problem of minimizing the total weighted completion time with ready times, weights and uniform processing times for all jobs. This layout is solvable in polynomial time, but the solution approach, a dynamic program, is characterized by a runtime of O(n18). The runtime is caused by a large number of states per iteration, which in turn are caused by the jobs’ ready times that make the sequencing more complex, and numerous options to schedule maintenance operations before/after or with a considered job.

An initial code implemented in Python by our research group can only solve very small instances in reasonable time. The applied testbed is provided by Drozdowski et al. (2016) and modified for the studied problem. Together with the hpc.bw project group at the Helmut Schmidt University, we aim to find ways to improve the code or even the formulation of the dynamic program by testing options to reduce the code, applying libraries to fasten the procedure, or finding ways to parallelize parts of the algorithm. Finally, we want to gain learning effects for our next projects dealing with similar scheduling problems.

Optimization of an IGA Code in MATLAB for parallel computing

Dr.Ing. Georgios Michaloudis (Professorship Structural Analysis, UniBw M)

Goal of this project within the framework of hpc.bw is the optimization of the performance of an In-House MATLAB Code, which is being developed from our researchers. The general area of application of the Code is Computational Structural Mechanics by deploying the Isogeometric Analysis framework. Novel computational methods are being permanently deployed aiming at investigations in the fields of Fracture Mechanics and of the analysis of shell and membrane structures. The Code can perform static as well as dynamic analyses. Weiterlesen

Monte Carlo simulations of real fluids

Univ.-Prof. Dr.Ing. Karsten Meier (Professorship Thermodynamics, HSU)

Thermodynamic properties of gases and liquids over wide ranges of temperature and pressure can be determined by molecular-dynamics or Monte Carlo simulations. In such simulations, the interactions between molecules are almost exclusively modeled by simple effective pair potentials, such as the Lennard-Jones potential. With such an approach, accurate predictions are only possible in small regions of temperature and pressure.  Our research aims at the very accurate prediction of thermodynamic properties and phase equilibria of noble gases, nitrogen, and carbon dioxide using sophisticated ab initio potentials, which are derived from quantum-chemical calculations of the interaction energies, by semi-classical Monte Carlo simulations. Weiterlesen

Enabling High-Throughput Studies of Reactive Materials

Christopher Lange (Department of Mechanical Engineering, HSU)

The project aims at establishing an optimized simulation framework for the atomistic modeling of impact-induced chemical reactions in new weapon materials. In our mechano-chemical research we combine atomic and molecular dynamics simulations. Strain-dependent reaction paths are determined with static ab initio simulations and coupled with impact-induced strains found by molecular dynamics. Together with the experts of hpc.bw we want to increase our research efficiency through better connections of our different simulations, their optimization for the HPC environment, and improved data management. Weiterlesen

benEFIT- Numerical simulation of non-destructive testing in concrete

Fabian Dethof (Professorship Engineering Materials and Building Preservation, HSU)

In order to assess the current state of concrete structures like a buildings or bridges without causing any damage, different non-destructive testing (NDT) methods can be performed. Some of those methods are based on elastic wave propagation like the ultrasonic testing (UT) or the impact echo (IE) method. The impact echo method is used to measure the resonance frequency of a plate-like structure. Therefore, a mechanical impact is generated (often manually) and the resulting waveform is measured at the structure’s surface just a few centimetres away from the impact point. Weiterlesen

The following projects are supported by hpc.bw:

  • C-STAR Electric Propulsion Demonstrator Multiphysics Modeling (Maximilian Maigler, UniBw M)
  • DigiTaKS* Learning behavior of students in dealing with digital media and tools (Prof. Dr. Sabine Schmidt-Lauff and Dr. Therese Rosemann, HSU)

C-STAR Electric Propulsion Demonstrator Multiphysics Modeling

Maximilian Maigler (Professorship for Plasma Technology and basics of Electrical Engineering, UniBw M)

The C-STAR experimental thruster is a novel electric propulsion system developed at the Laboratory of Plasma Technology of the UniBw Munich. It is a vacuum arc thruster, which employs a strong direct-current arc to ionize a propellant traversing the arc. Using permanent magnets, the heavy ions are accelerated and produce thrust. The specific impulse and therefore efficiency can be up to 30 times higher compared to conventional chemical thrusters. In order to estimate the performance and lifetime of this experimental thruster, we setup and carry out holistic multiphysics simulations of the entire system. The C-STAR thruster will be operating on board the SeRanis satellite and tested for orbital attitude control.

DigiTaKS* Learning behavior of students in dealing with digital media and tools

Prof. Dr. Sabine Schmidt-Lauff and Dr. Therese Rosemann (Professorship for continuing education and lifelong learning, HSU)

The cooperation of the project Hpc.bw (Competence Platform for Software Efficiency and Supercomputing) and the research project DigiTaKS* (Digital competences for study and work – development of a model for transformative digital competences for students) aims at the implementation of an automated, individualized result feedback on (digital) learning and usage behavior. As part of the DigiTaKS* project, approximately 400 students will be provided with hard- and software. Accordingly, the learning behavior in dealing with digital media and tools of students is collected via a longitudinal diary study over 10 days. Hpc.bw provides software to facilitate individualized, automated results feedback for students of the DigiTaKS* project. The students of the research project will be provided with an individualized learning and usage profile in dealing with digital media and tools. The individual feedback promotes reflexivity and sensitivity for one’s own learning behavior, so that the students recognize specifics of individual learning and patterns of use (habits) studying and working with digital media and tools. By comparing the individual results of the students with the characteristics of the entire group of students in the DigiTaKS* project, they receive feedback on their own digital learning and usage behavior as well as habits. The results of the individual feedback can increase students‘ sensitivity to individual learning and development needs. The cooperation is integrated in the project-related habilitation of Dr. Therese Rosemann.

Do you have any questions about the project? You can reach the project team at

[email protected]

 

dtec.bw_EU
HSU

Letzte Änderung: 10. Juli 2023