HLRS Virtual Booth at SC21

The SC21 Conference – the world’s leading HPC event – will take place from November 14-19, 2021 as a hybrid event, both onsite in St. Louis, MO, USA and online. Due to travel difficulties resulting from the ongoing global COVID-19 pandemic, the High-Performance Computing Center Stuttgart (HLRS) will be participating remotely through a virtual booth together with the HiDALGO Center of Excellence. Although we regret not being able to be there personally, we invite you to join us.

The following schedule details HLRS- and HiDALGO-related presentations that will be given at SC. These talks will be streamed online via Zoom.

Visit our booth on the SC21 HUBB!

Click here to visit the SC21 website and register. 

Schedule

Please note that times are given in Central Standard Time (St. Louis, CST) and Central European Time (CET).

Tuesday, November 16, 2021

12:00 - 12:45 (CST)
19:00 - 19:45 (CET)
Forecasting intensive care unit demand during the COVID-19 pandemic: 10 months result review and current developments
13:00 - 13:30 (CST)
20:00 - 20:30 (CET)
Covid-19 pandemic data analysis and the urge for federated learning
Wednesday, November 17, 2021 (HiDALGO CoE)
09:00 - 09:15 (CST)
16:00 - 16:15 (CET)
Introducing how HPC can address global challenges
09:15 - 09:45 (CST)
16:15 - 16:45 (CET)
Case study: Coupling of multiscale forced displacement and weather simulations — the South Sudan conflict
09:45 - 10:15 (CST)
16:45 - 17:15 (CET)
Case study: Introducing FACS — a hyperlocal and parallel agent-based COVID-19 simulation code
10:15 - 10:45 (CST)
17:15 - 17:45 (CET)
Case study: Simulation urban air pollution
10:45 - 11:15 (CST)
17:45 - 18:15 (CET)
Case study: Spectral methods to compute eigenvalue histograms in the context of social network analysis
11:30 - 12:00 (CST)
18:30 - 19:00 (CET)
Visualization of air pollution using digital urban twins
12:00 - 12:30 (CST)
19:00 - 19:30 (CET)
Big data and supercomputers: Utilizing high performance data analytics in an HPC setup
12:30 - 13:00 (CST)
19:30 - 20:00 (CET)
CPU and GPU optimization technics implemented in HiDALGO's use case applications
Thursday, November 18, 2021
09:30 - 10:15 (CST)
16:30 - 17:15 (CET)
Addressing SME challenges through the FF4EuroHPC experiments
10:30 - 11:15 (CST)
17:30 - 18:15 (CET)
Strengthening HPC competences in Europe in the Exa-Era
11:15 - 12:00 (CST)
18:15 - 19:00 (CET)
Quantum algorithms complexity for linear systems of engineering (SEQUOIA)
12:00 - 13:00 (CST)
19:00 - 20:00 (CET)
Scientific visualization: Combining in-situ, remote rendering and virtual/augmented reality

Forecasting intensive care unit demand during the COVID-19 pandemic: 10 months result review and current developments

Forecasting intensive care unit demand during the COVID-19 pandemic: 10 months result review and current developments

November 16, 2021: 12:00 - 12:45 (CST)

The COVID-19 pandemic poses the risk of overburdening health care systems. However, especially the NPIs like partial lockdowns can have negative effects on the economy and social life. Thus a correct and permanently updated understanding of the impact of the pandemic on ICU demand is vital for political decision-making. We present our implementation of a spatial age-structured microsimulation model of the COVID-19 pandemic that relies on an extended SEIR framework along with its automatic forecast procedure for ICU coverage. We will review the results delivered during 10 month of production and present current model developments and performance improvements.

Speaker: Dr. Ralf Schneider (HLRS)

Covid-19 pandemic data analysis and the urge for federated learning

Covid-19 pandemic data analysis and the urge for federated learning

November 16, 2021: 13:00 - 13:30 (CST)

In the project ORCHESTRA ("Connecting European Cohorts to Increase Common and effective Response to SARS-CoV-2 Pandemic") HLRS, together with 27 centers for public health and high-performance computing from 15 countries in Europe, Africa, South America, and Asia, will help develop a data infrastructure for collecting and analyzing COVID-19 patient data from across Europe and other parts of the world.

HLRS will help to build a federated research architecture for cohort analysis based on three layers: national data providers, national hubs, and the centralized ORCHESTRA portal. A national hub can be envisioned as a system for cohort data and metadata storage. The Italian HPC Center CINECA will summarize a state of the art of the GDPR-compliant Data Infrastructures within European Supercomputing centers involved in ORCHESTRA. The national hubs will be complemented with tools to enable federated learning, which is an approach to jointly analyze data when it can’t be moved to other locations due to legal regulations, security restrictions or performance considerations. Within the project, DataShield will be used as the main technology to implement secure federated learning.

Speakers: Dr. Björn Schembera (HLRS), Dr. Gabriella Scipione (CINECA)

Introducing how HPC can address global challenges

Introducing how HPC can address global challenges

November 17, 2021: 09:00 - 09:15 (CST)

As part of the EU-funded project HiDALGO, we develop novel methods, algorithms and software for HPC and HPDA to accurately model and simulate the complex processes, which arise in connection with major global challenges.

Speaker: Francisco Javier Nieto de Santos (Atos)

Case study: Coupling of multiscale forced displacement and weather simulations — the South Sudan conflict

Case study: Coupling of multiscale forced displacement and weather simulations — the South Sudan conflict

November 17, 2021: 09:15 - 09:45 (CST)

We propose a multiscale forced displacement simulation approach that can assist organisations with the allocation of humanitarian resources. To provide more accurate simulations, we consider perceived levels of safety and road accessibility influenced by the level of precipitation and river discharge. Most studies in this area have only considered the effects of climate on forced displacement, but there are no studies on the impact of weather on people’s decision to flee or travel. Aiming to fill this gap, we investigate precipitation and river discharge levels affecting the movement speed of forcibly displaced people and their decision to remain in their current location or traverse through other routes. Specifically, we analyse the South Sudan conflict between 2016-2017 and evaluate the coupled multiscale and single-scale simulation results by comparing the total validation error, execution time and coupling overhead.

Speaker: Diana Suleimenova (Brunel University)

Case study: Introducing FACS — a hyperlocal and parallel agent-based COVID-19 simulation code

Case study: Introducing FACS — a hyperlocal and parallel agent-based COVID-19 simulation code

November 17, 2021: 09:45 - 10:15 (CST)

The Flu And Coronavirus Simulator, or FACS, is an agent-based simulation code that mimics the spread of COVID-19 in a local area, such as a municipality or a borough in a large city that serves as a catchment area for one or a few hospitals. Using FACS, we are able to (a) derive virtual human households from geospatial (OpenStreetMap) and demographic (ONS) data sources and (b) calculate how the people in these households infect each other as they visit a wide array of nearby locations. The code is unique in that it has an explicit sub-national/local scope, extracts its location graphs directly from geospatial data sources, and that FACS users have shown to be able to construct and execute new simulations in a matter of days.

Speaker: Derek Groen (Brunel University)

Case study: Simulation urban air pollution

Case study: Simulation urban air pollution

November 17, 2021: 10:15 - 10:45 (CST)

The aim of the urban air pollution application is to create cleaner air in cities by using high performance computing (HPC) and mathematical technologies. To this end, this research will provide policy makers and society with an easy-to-use computational tool as a service that accurately and quickly forecasts air pollution in cities with very high resolution. Furthermore, a traffic control system will be developed as well to minimize air pollution while considering traffic flow constraints.

Speaker: Zoltán Horváth (Széchenyi István Egyetem)

Case study: Spectral methods to compute eigenvalue histograms in the context of social network analysis

Case study: Spectral methods to compute eigenvalue histograms in the context of social network analysis

November 17, 2021: 10:45 - 11:15 (CST)

Social networks are omnipresent in our daily life. They are used for communication and information dissemination, which highly influences our behavior. One of the main fields in social network analysis is the study of the topological and algorithmic properties of the underlying graphs, and the design of proper random graph models to describe these networks. Once a random graph model has been generated, it is crucial to compare its properties to the corresponding real-world networks. For this, the comparison of eigenvalue histograms is essential.

Speaker: Robert Elsässer (Paris Lodron Universität Salzburg)

Visualization of air pollution using digital urban twins

Visualization of air pollution using digital urban twins

November 17, 2021: 11:30 - 12:00 (CST)

Digital urban twins can facilitate the understanding of complex, interdisciplinary processes in cities and regions. Within the HiDALGO project, a digital twin for the city of Stuttgart, Germany was employed for supporting the analysis of urban air pollution simulations. From the urban air pollution use case, estimates for the distribution of pollutants such as nitric oxides were obtained. In combination with further data sets these estimates can be studied in the frame of a digital twin. Visualization in Virtual Reality allows low-threshold access in an immersive, interactive manner. The presentation will be streamed live from the CAVE.

Speaker: Dennis Grieger, Dr. Fabian Dembski, Marko Djuric (HLRS)

Big data and supercomputers: Utilizing high performance data analytics in an HPC setup

Big data and supercomputers: Utilizing high performance data analytics in an HPC setup

November 17, 2021: 12:00 - 12:30 (CST)

In this presentation we will discuss the role of high performance data analytics (HPDA) in an HPC setup. Our goal is to showcase how HPDA has been utilized and leveraged to obtain insights into the simulation outcomes of the HPC processing tasks of HIDALGO, present some of the challenges we faced during the design, implementation and deployment of the HPDA methods and discuss lessons learned in the process.

Speaker: Nicolaos Chalvantzis (CSLab, National Technical University of Athens)

CPU and GPU optimization technics implemented in HiDALGO's use case applications

CPU and GPU optimization technics implemented in HiDALGO's use case applications

November 17, 2021: 12:30 - 13:00 (CST)

The Global Challenges applications are highly demanding in terms of the amount of processed data. This implicates that the application design must demonstrate sufficient resiliency to operate efficiently on incoming data on a given timeframe, taking into account manifold data sources and their particular characteristics. The first phase of yield testing is assesment of its performace while in the next step the focus is concentrated on excellence in application design and optimisation by examining algorithms, programming models, techniques and implementation methods to enable near-linear scalability for both compute-intensive and dataintensive applications. The achievements will be presented on the basis of the results of the analysis of pilot applications coming from "Urban Air Pollution" and "Social Networks" use cases and conducted on both CPU and GPU units.

Speaker: Marcin Lawenda (Poznan University of Technology)

Addressing SME challenges through the FF4EuroHPC experiments

Copyright: Arctur
Copyright: Arctur
Addressing SME challenges through the FF4EuroHPC experiments

November 18, 2021: 09:30 - 10:15 (CST)

We are in the middle of a digital revolution being driven by data and intelligence, including technologies such as high-performance computing (HPC), artificial intelligence (AI), high-performance data analytics (HPDA) and machine learning (ML). Many SMEs can benefit from using high-performance computing and data analytics techniques in their business but are put off because of worries about the complexity, the lack of skilled personnel, and the cost. FF4EuroHPC is a European initiative that helps facilitate access to all high-performance computing-related technologies for SMEs and thus increases the innovation potential of European industry. Through two open calls, innovative HPC experiments addressing SME business challenges will be selected. The first wave of experiments is running and working to solve a range of business challenges through the use of advanced HPC services. The second open call closed recently and the evaluation of proposals addressing a wide variety of industrial sectors from across Europe is currently ongoing so that the second wave of experiments can get started early in 2022. Dr. Lonsdale will present the activities of the FF4EuroHPC project, as well as a number of SME success stories illustrating benefits of using HPC to solve business challenges.

Speaker: Dr. Guy Lonsdale (scapos AG)

Strengthening HPC competences in Europe in the Exa-Era

Strengthening HPC competences in Europe in the Exa-Era

November 18, 2021: 10:30 - 11:15 (CST)

This talk will provide an overview about activities being led by EuroHPC and in other contexts to realize the European HPC strategy, with a focus on the EuroCC and CASTIEL projects. The main emphasis will be on activity surrounding the implementation of National Competence Centres for high-performance computing in 33 nations and their strategic alignment on the European Level.

Speaker: Dr. Natalie Lewandowski (HLRS)

Quantum algorithms complexity for linear systems of engineering (SEQUOIA)

Quantum algorithms complexity for linear systems of engineering (SEQUOIA)

November 18, 2021: 11:15 - 12:00 (CST)

Modern and future numerical simulations require high-resolution models with an almost unlimited number of unknowns. In many cases, solving the underlying large sparse linear systems with up to billions of the unknowns, takes up the most time and consumes a lot of energy. Working together in the SEQUOIA project with the Fraunhofer Institute for Industrial Engineering (Fraunhofer IAO) and five additional partners, HLRS is investigating the capability of known quantum algorithms, such as Harrow, Hassidim Lloyd (HHL) and Variational Quantum Linear Solver (VQLS) to significantly reduce the complexity of solving large sparse linear systems. In particular, the presentation will address the challenges and consequences of Noisy Intermediate-Scale Quantum (NISQ) technology.

Speakers: Dmitry Khabi (HLRS)

Scientific visualization - Combining in-situ, remote rendering and virtual/augmented reality

Copyright: HLRS
Copyright: HLRS
Scientific visualization - Combining in-situ, remote rendering and virtual/augmented reality

November 18, 2021: 12:00 - 13:00 (CST)

Visualization of scientific data provokes challenges in terms of data size, complexity, and availability. Approaches such as in-situ visualization, parallel and remote rendering can be used to overcome such barriers. In this demonstration, state-of-the-art implementations of these approaches will be presented. Moreover, Virtual and Augmented Reality is leveraged enabling real-time analysis of heterogenous HPC data. Several examples will be shown live from a CAVE addressing inter alia applications from tsunami research or architecture. Besides, examples from Augmented Reality will be presented.

Speakers: Dennis Grieger, Dr. Fabian Dembski, Marko Djuric (HLRS)