G14
HPC 3

Back to overview

17:35
conference time (CEST, Berlin)
Creating Connections: Enabling High Performance Computing for Industry through a Data Exchange & Workflow Platform
27/10/2021 17:35 conference time (CEST, Berlin)
Room: G
J. Schüssler, J. Grimm (SSC-Services GmbH, DEU)
J. Schüssler, J. Grimm (SSC-Services GmbH, DEU)
SSC-Services GmbH (SSC) is developing a secure data exchange and transfer platform within the Centre of Excellence in Engineering (EXCELLERAT) to ease the use of High-Performance Computing (HPC) resources and make data transfer more efficient. Nowadays, organisations and industrial partners face various issues while dealing with HPC calculations, HPC in general or even the access to HPC resources. In many cases, calculations are too complex and potential users do not have the required expertise to benefit from HPC technologies without support. SSC deals with this topic while this challenge has been taken on. The developed Data Exchange and Workflow Portal will be able to simplify or even eliminate these obstacles. The new platform enables users to access easily the two HLRS clusters, Hawk and Vulcan, from any authorised device and to run their simulations remotely. In the future, further data centres will also be connected to the platform. With the help of the platform, users do not have to use command-line tools to manage their simulations, run them and retrieve the results, which requires a lot of training or support effort on the part of the HPC, especially for inexperienced users. The added value for the industrial customer is for example the reduction of HPC complexity due to web frontend which leads to a higher HPC customer retention due to less complex HPC environment. Furthermore, calculations can be started from anywhere with a secure connection. There are also time and cost savings due to a high degree of automation that simplifies the process chain. In cooperation with various pilot partners from the industry, the platform prototype is undergoing various tests regarding its suitability, starting with the German Federal Institute for Population Research as the first learning project with real productive use. All user requirements and feedbacks are incorporated into further development and optimization to offer the greatest possible added value for future users from different areas like automotive, aerospace, energy and manufacturing areas. The presentation will briefly introduce the EXCELLERAT project, address the challenges faced by industrial HPC users and present solutions through different use cases. Dialogs with experts and potential pilot partners will also be sought to capture their needs and to be able to include the required competencies in the further course of the project and in the further development of the platform.
Data Transfer, Data Management, Data Reduction, Automatisation, Simplification
17:55
conference time (CEST, Berlin)
Algebraic Multigrid (AMG) for Large-scale CFD Simulations
27/10/2021 17:55 conference time (CEST, Berlin)
Room: G
B. Metsch, H-J. Plum (Fraunhofer SCAI, DEU)
B. Metsch, H-J. Plum (Fraunhofer SCAI, DEU)
Linear solvers form the inner core of many simulations in science and engineering. For example, in computational fluid dynamics (CFD), the numerical solution of the Navier-Stokes equations require the solution of a large, sparse linear system to obtain the pressure solution and, in consequence, ensure the consistency of the velocity field. While the discretization of the partial differential equations involved is a local operation and thus can be parallelized easily, the linear solution is a global operation and thus imposes a significant challenge for the overall scalability of the simulation. Algebraic multigrid (AMG) methods provide optimal iterative linear solvers for a wide class of problems, i.e. they show a linear scaling with respect to N in both compute time and memory requirements. They automatically construct a hierarchy of linear systems to adequately deal with the different frequencies of the underlying problem. To this end, a so-called setup phase is carried out before the iteration starts. In this talk, we present the parallel scalability of the algebraic multigrid solver SAMG inside an OpenFOAM CFD solver. We focus on how to improve the speed-up (i.e. strong scaling) for large-scale simulations (10-100Ms of degrees of freedom, 10000s of time steps) on industrially relevant processor numbers. To improve the parallel performance, we combine several components: First, we employ parallel AMG algorithms that lower the communication overhead and thus improve the scalability[1,2]. Second, we only solve as accurate as needed: We reduce the number of iterations where possible, and use single precision arithmetic if the accuracy required allows so. Furthermore, we have developed a steering mechanism that carefully inspects how often the AMG setup phase needs to be carried out. Finally, for selected examples we have carried out a compilation optimization process to gain further speed-up. [1] H. De Sterck, U.M. Yang, and J.J. Heys, Reducing Complexity in Parallel Algebraic Multigrid Preconditioners, SIAM J. on Matrix Analysis and Applications, 27 (2006), pp. 1019-1039. UCRL-JRNL-206780. [2] R.D. Falgout and J.B. Schroder, Non-Galerkin Coarse Grids for Algebraic Multigrid,SIAM J. Sci. Comput., 36 (2014), pp. C309-C334. LLNL-JRNL-641635.
computational fluid dynamics, algebraic multigrid, parallel computing, high performance computing
18:15
conference time (CEST, Berlin)
GPU Accelerated Radiative View Factor Calculations: Applications in Headlamp Design and Thermal Analysis
27/10/2021 18:15 conference time (CEST, Berlin)
Room: G
M. Kenward, M.T. Nguyen, M. Saghir, N. Hayi-Slayman, P. Wen, Z. Zarei (Maya HTT, CAN)
M. Kenward, M.T. Nguyen, M. Saghir, N. Hayi-Slayman, P. Wen, Z. Zarei (Maya HTT, CAN)
The thermal analysis of headlamps and related engineering applications is a key step to ensure their safety, longevity, and can be a driving factor in their overall design process. A comprehensive understanding of the thermal physics in these applications is crucial to analysts and designers. The role of simulation in headlamp design is a fundamental part of this process. The time spent by analysts to evaluate the thermal characteristics of a product design has a fundamental impact on the engineering of headlamps and the entire design process. A significant bottleneck in the thermal analysis is the computationally costly calculation of radiative view factors, which dictate the radiative conductances in headlamp thermal physics. Historically, several approaches have been used to tackle this problem. However, existing methods do not tend to yield the performance and scalability required for optimally efficient thermal analysis. Considering this, we present a GPU-based tool for radiative view factor calculations implemented in the CUDA programming language. We begin by outlining the existing approaches and their pitfalls, including the Hemicube method and Monte Carlo ray tracing methods using conventional CPU-based parallelism. We then discuss our GPU based Monte Carlo view factor calculation method, which is designed to harness the embarrassingly parallel nature of view factor (ray tracing) calculations and the large number of compute cores on GPUs. We show that our implementation is accurate, highly parallel, has been validated against known results, and orders of magnitude faster than existing CPU-based methods run in distributed parallel environments. We then focus on the specific case study of headlamp design and the improvements in the speed of view factor calculations. The new method has the promise to fundamentally change the approach and time needed for design iterations in some engineering applications. We conclude with an outlook for applications in other industries in the near term.
GPU, heat transfer, simulation, modeling
18:35
conference time (CEST, Berlin)
Democratizing a Highly Parallelized CFD Model to Facilitate the Study of Formation Flight
27/10/2021 18:35 conference time (CEST, Berlin)
Room: G
A. Perez, Dr. A.Tejada-Martinez (University of South Florida, USA ); P. Pennington (EASA, Inc. USA)
A. Perez, Dr. A.Tejada-Martinez (University of South Florida, USA ); P. Pennington (EASA, Inc. USA)
"Formation flight has been known to provide performance benefits to migrating birds for over a century. Very optimistic estimates for the range extension experienced by migrating birds made in the 70’s prompted interest in implementing formation flight for range extension in civil and military aviation. Some successful implementations include the NASA AFF program and the DARPA/AFRL SAVE program where nearly 25% and 15% maximum drag reduction was achieved, respectively. In the literature, Computational Fluid Dynamics (CFD) has been applied to studying this potentially useful approach beyond what is feasibly possible through physical experimentation. However, due to the additional complexity associated with the parametric nature of this approach, CFD has been limited to lower order models such as vortex lattice methods, 3D-Euler simulation, and 3D Reynolds-Averaged Navier-Stokes (RANS) simulation. Indeed, this complexity is not strictly computational, as software limitations, preprocessing and postprocessing make using a more robust CFD methodology such as Large Eddy Simulation (LES) impractical. Therefore, we seek to address these challenges through a suite of CFD simulations using an open source, highly parallelized, CFD software capable of LES along with a “low” code platform (EASA). We employ LES to determine the aerodynamic forces felt by two and three generic blended wing Unmanned Aerial Vehicles (UAVs) in cruise conditions as a function of their relative positioning and various flight parameters. Meanwhile the “low” code platform (EASA) is used to allow the rapid batching of different cases, solution monitoring and rapid post-processing. The use of a “low” code model deployment platform represents a novel means of democratizing not only a GUI-less, open-source engineering simulation software but also a means to democratize access to high-performance computing in general. The enhanced usability and productivity will be showcased. Finally, these results will be compared to a commercial CFD code such as ANSYS Fluent, along with cost and feature analyses between the two methods."
Democratization, parallelized, CFD, formation flight, efficiency
×

[TITLE]

[LISTING

[ABSTRACT]

[DATE]

[ROOM]

[KEYWORDS]