Artikel in Tagungsband INPROC-2020-16

Bibliograph.
Daten
Hirschmann, Steffen; Kronenburg, Andreas; Glass, Colin W.; Pflüger, Dirk: Load-Balancing for Large-Scale Soot Particle Agglomeration Simulations.
In: Foster, Ian (Hrsg); Joubert, Gerhard R. (Hrsg); Kucera, Ludek (Hrsg); Nagel, Wolfgang E. (Hrsg); Peters, Frans (Hrsg): Parallel Computing: Technology Trends.
Universität Stuttgart, Fakultät Informatik, Elektrotechnik und Informationstechnik.
Advances in Parallel Computing; 36, S. 147-156, englisch.
IOS Press, März 2020.
DOI: 10.3233/APC200035.
Artikel in Tagungsband (Konferenz-Beitrag).
KörperschaftInternational Conference on Parallel Computing
CR-Klassif.G.0 (Mathematics of Computing General)
Kurzfassung

In this work, we combine several previous efforts to simulate a large-scale soot particle agglomeration with a dynamic, multi-scale turbulent background flow field. We build upon previous simulations which include 3.2 million particles and implement load-balancing into the used simulation software as well as tests of the load-balancing mechanisms on this scenario. We increase the simulation to 109.85 million particles, superpose a dynamically changing multi-scale background flow field and use our software enhancements to the molecular dynamics software ESPResSo to simulate this on a Cray XC40 supercomputer. To verify that our setup reproduces essential physics we scale the influence of the flow field down to make the scenario mostly homogeneous on the subdomain scale. Finally, we show that even on the homogeneous version of this soot particle agglomeration simulation, load-balancing still pays off.

Volltext und
andere Links
PDF (344061 Bytes)
Abteilung(en)Universität Stuttgart, Institut für Parallele und Verteilte Systeme, Simulation großer Systeme
Eingabedatum16. April 2020
   Publ. Institut   Publ. Informatik