This event has passed.

35th Workshop on Sustained Simulation Performance

Name: 35th Workshop on Sustained Simulation Performance
Start: 2023-04-13T00:00:00+01:00
End: 2023-04-14T23:59:59+01:00
Location: HLRS

April 13, 2023 - April 14, 2023

Agenda

All times are given in Central European Summer Time (CEST).

The registration is closed.

Thursday, April 13, 2022
10:00 – 10:15	Welcome & Introduction Michael Resch, HLRS, University of Stuttgart
10:15 – 10:45	Lessons Learned from A Quantum-Annealing Assisted HPC R&D Project Hiroaki Kobayashi, Tohoku University
10:45 – 11:15	NEC’s Quantum Computing Strategy, Technology, and Use Cases Shintaro Momose, NEC Corporation This presentation consists of two parts, discussing SX-Aurora TSUBASA vector supercomputer and introducing simulated annealer working on SX-Aurora TSUBASA called Aurora Vector Annealing. The first half of the presentation shows the vector architecture of SX-Aurora TSUBASA, especially its latest vector processors having the highest-level memory bandwidth. Sustained performance and power efficiency are also discussed, as well as NEC’s future plans and roadmap. The second half of the presentation shows NEC’s quantum computing strategies and their products to provide higher sustained performance in the annealing/optimization fields. NEC developed the Aurora Vector Annealing as a simulated annealer and has a strong business relationship with D-Wave providing a quantum annealer. NEC aims at solving various social issues by using the quantum/simulated annealing technologies and by developing a hybrid platform with supercomputer and quantum/simulated annealer to provide much higher sustained performance.
11:15 – 11:45	Towards Science DMZ based on Accelerated ONION using DTN Susumu Date, Osaka University The speaker introduces what is happening in Osaka University towards the promotion and advancement of Data-driven Scientific Research. In this talk, the experience of using DTN is first introduced and then a future direction of compute infrastructure composed of supercomputers( SQUID and OCTOPUS ) and data infrastructure is explained based on the experience.
11:45 – 13:15	Lunch
13:15 – 13:45	SPH-EXA: A Framework for Scalable, Flexible, and Extensible Astrophysical and Cosmological Simulations – Slides Florina M. Ciorba, Department of Mathematics and Computer Science, University of Basel SPH-EXA is a highly scalable and extensible simulation framework for astrophysical and cosmological simulations. It is codesigned by computational (astrophysicists and cosmologists) and computer scientists (high-performance computing) to achieve scientific advances in astrophysics, cosmology, and high-performance computing, for highly scalable simulations using Smoothed Particle Hydrodynamics. SPH-EXA includes highly optimized and parallelized hydrodynamics and gravity solvers. It supports new particle types and particle data fields that can be combined with custom-made, simulation-derived observable properties and in-situ data analysis. It requires minimal software dependencies, provides scalable parallelization and communication support, and portability with optimizations for recent CPUs and GPUs. This design relieves potential users from architectural specifics and performance concerns for performing and scaling up simulations with SPH-EXA. This talk will describe the SPH-EXA framework, its approach to domain decomposition, parallelization, gravity and hydrodynamics solvers, and their flexible use as pluggable components. A scalable initial conditions generator, test cases and simulations will also be presented. SPH-EXA is extensible with additional physical effects through the use of propagators, whereby the pluggable framework components are easily customized or implemented from scratch using the provided building blocks in an abstract, efficient, and scalable way.
13:45 – 14:15	Prediction and Mitigation of Aeroacoustic Noise on HPC Systems Matthias Meinke, Ansgar Niemöller, Miro Gondrum, Zhe Yang, Wolfgang Schröder, Institute of Aerodynamics, RWTH Aachen University, Germany
14:15 – 14:45	Real-time flood inudndation simulation on SX-Aurora TSUBASA Hiroyuki Takizawa, Yoichi Shimomura, Akihiro Musa, Yoshihiko Sato, Atuhiko Konja, Guoqing Cui, Rei Aoyagi, and Keichi Takahashi, Tohoku University A real-time flood inundation simulation based on the Rainfall-Runoff Inundation (RRI) model is memory-intensive, and SX-Aurora TSUBASA is hence promising to efficiently execute the simulation in time due to high sustained memory bandwidth provided by vector processors. This talk will report that the real-time simulation has successfully been migrated and optimized for SX-Aurora TSUBASA. Assuming a shared computing system such as Supercomputer AOBA installed at Tohoku University, a resource demand estimation method is developed to minimize the amount of shared computing resources used for prediction in order to reduce the impact on other users sharing the system. The evaluation results show that SX-Aurora TSUBASA with only 32 cores can meet the real-time simulation requirement of simulating 7-hour flood inundation for the Tohoku region of Japan within 20 minutes, and also the resource demand estimation method can adaptively adjust the computing resource amount used for the real-time simulation.
14:45 – 15:15	Break
15:15 – 15:45	Competence Centres and Centres of Excellence within the European Strategy – Slides Bastian Koller, HLRS, University of Stuttgart
15:45 – 16:15	Power capping in high performance computing – experiences and prospects Pawel Czarnul, Adam Krzywaniak and Jerzy Proficz, Gdańsk University of TechnologyIn this work we investigate usage, limitations and prospects of power capping in high performance computing (HPC). Specifically, we discuss APIs for modern CPUs, GPUs and present, as an illustration, new unpublished data showing performance and energy characteristics for selected parallel OpenMP applications under power caps executed on a dual socket Intel Skylake-X system. These APIs can be used within algorithms: deriving configurations for which selected performance-energy goals (such as EDP, EDS) are optimized for non-trivial i.e. non-default power caps; and also allowing minimization of execution times under power caps. We discuss various factors of interest in the future in power capping aware HPC such as other metrics not considered so far, applicability and accuracy of measurement methods: using filters, hardware vs software methods, conditions and use cases for particular methods. We describe future works and areas including scenarios that can benefit from power capping in HPC.
16:15 – 16:45	Containerization for DLR HPC applications Sabine Roller, Deutsches Zentrum für Luft- und Raumfahrt, Institut für Softwaremethoden zur Produkt-Virtualisierung
18:30	Dinner
Friday, April 14, 2023
09:00 – 09:45	Keynote – Sustaining Simulation Performance in the US Exascale Computing Project – Slides Hartwig Anzt , University of Tennessee, Knoxville The US Exascale Computing Project (ECP) has the goal to deliver a capable exascale computing ecosystem to provide breakthrough modeling and simulation solutions to address the most critical challenges in scientific discovery, energy assurance, economic competitiveness, and national security. This requires providing scientific computing applications with a software stack that allows them to perform on the leadership supercomputers. In this talk, we discuss the impact of the ECP hardware landscape on software design and how the Ginkgo math library responds to the ECP application requirements and helps to achieve the simulation performance goals.
09:45 – 10:15	HPC and AI at HLRS – Slides Michael Resch, HLRS, University of Stuttgart HLRS was created to support users with High-Performance Computing. Over the last years, however, Artificial Intelligence has become an important topic. In this talk we look into the changes that AI drives in computer simulation. We will explore the needs of AI users. Furthermore, we will have a first look at the needs of traditional computer simulation and t the potential of AI for such simulations.
10:15 – 11:00	Break
11:00 – 11:30	Towards Building a Digital Twin of Job Scheduling Tatsuyoshi Ohmura, NEC Corporation A job scheduler, which is the core component of the HPC system, has several parameters. Tuning these parameters can improve the efficiency of system operation, but this requires knowledge and experience and places a burden on the system operator. To reduce this burden, we are studying the digital twin of the job scheduler. To realize the digital twin, we have developed a job scheduler simulator. We demonstrate examples of the use of the simulator.
11:30 – 12:00	Scalable Cluster Administration with LXC³ – Slides Erich Focht, NEC Corporation The LXC^3 Cluster Command and Control tools have evolved at NEC Germany since two decades, going through various changes to adapt to continuously changing requirements. The talk discusses the design choices of LXC^3-neo which runs as a pool of micro-services orchestrated by docker swarm, its scalability and limitations seen at customer sites. The most recent developments move the cluster management stack even closer to methods used in cloud systems management, simplifying node image handling and management network setup.
12:00 – 13:30	Lunch
13:30 – 14:00	A feasibility study of quantum annealing for the next-generation computing infrastructure – Slides Kazuhiko Komatsu, Tohoku University This presentation introduces a new project, feasibility study of quantum computing for the next-generation computing infrastructure, and shows an early evaluation of annealing machines.
14:00 – 14:30	Energy Efficiency and Renewable Energy for Distributed High-Performance Computing Christoph Niethammer, HLRS, University of Stuttgart
14:30 – 15:00	A new framework for calibrating COVID-19 SEIR models with spatial-/time-varying coefficients using genetic and sliding window algorithms – Slides Huan Zhou, HLRS, University of Stuttgart A susceptible-exposed-infected-removed (SEIR) model assumes spatial-/time-varying coefficients to model the effect of non-pharmaceutical interventions (NPIs) on the regional and temporal distribution of COVID-19 disease epidemics. A significant challenge in using such models is their fast and accurate calibration to observed data from geo-referenced hospitalized data, i.e., efficient estimation of the spatial-/time-varying parameters. In this work, a new calibration framework is proposed towards optimizing the spatial-/time-varying parameters of SEIR model. We also devise a method for combing the overlapping sliding window technique (OSW) with a genetic algorithm (GA) calibration routine to automatically search the segmented parameter space. Parallelized GA is used to reduce the computational burden. Our framework abstracts the implementation complexity of the method away from the user. It provides high-level APIs for setting up a customized calibration system and consuming the optimized values of parameters. We evaluated the application of our method on the calibration of a spatial age-structured microsimulation model (CoSMic) using a single objective function that comprises observed COVID-19-related ICU demand. The results reflect the effectiveness of the proposed method towards estimating the parameters in a changing environment.
15:00	Open End

Details

Start:: April 13, 2023
End:: April 14, 2023

Organizer

: Mr. Johannes Gebert
Email: gebert@hlrs.de

Venue

: HLRS
: Nobelstraße 19
Stuttgart, Baden-Württemberg 70569 Germany + Google Map

Committee
Program Committee

Prof. Michael Resch, Stuttgart University, HLRS
Prof. Hiroaki Kobayashi, Tohoku University
Dr. Wolfgang Bez, NEC Deutschland GmbH, Division HPCE
Prof. Sabine Roller, German Research School for Simulation Sciences GmbH

Organizing Committee

Prof. Michael Resch, Stuttgart University, HLRS
Johannes Gebert, Stuttgart University, HLRS
Prof. Hiroaki Kobayashi, Tohoku University
Prof. Hiroyuki Takizawa, CyberScience Center, Tohoku University