Depicted is the system, the researchers used for their case study. A ripping fluid layer simulated with the large-scale molecular dynamic solver ls1 and visualized in situ with the visualization framework MegaMol.

November 29, 2019 /

SC19: Honorable Mention Award for VISUS researchers

Congratulations to our VISUS team!
[Picture: VISUS]

“The Impact of Work Distribution on In Situ Visualization: A Case Study” it the title of the latest publication by Tobias Rau, Patrick Gralka, Oliver Fernandes, Guido Reina, Steffen Frey ,and Thomas Ertl of the Visualization Research Center of the University of Stuttgart (VISUS). The researchers participated in the International Conference for High Performance Computing, Networking, Storage and Analysis (SC19, November 17 to 22, 2019) in Denver (USA) and received an Honorable Mention during the ISAV 2019 workshop, that took place on November 18, 2019.

Abstract

Large-scale computer simulations generate data at rates that necessitate visual analysis tools to run in situ. The distribution of work on and across nodes of a supercomputer is crucial to utilize compute resources as efficiently as possible. In this paper, we study two work distribution problems in the context of in situ visualization and jointly assess the performance impact of different variants. First, especially for simulations involving heterogeneous loads across their domain, dynamic load balancing can significantly reduce simulation run times. However, the adjustment of the domain partitioning associated with this also has a direct impact on visualization performance. The exact impact of this side effect is largely unclear a priori as generally different criteria are used for balancing simulation and visualization load. Second, on node level, the adequate allocation of threads to simulation or visualization tasks minimizes the performance drain of the simulation while also enabling timely visualization results. In our case study, we jointly study both work distribution aspects with the visualization framework MegaMol coupled in situ on node level to the molecular dynamics simulationls1 Mardyn on Stampede2 at TACC.

Links
Depicted is the system, the researchers used for their case study. A ripping fluid layer simulated with the large-scale molecular dynamic solver ls1 and visualized in situ with the visualization framework MegaMol.
Depicted is the system, the researchers used for their case study. A ripping fluid layer simulated with the large-scale molecular dynamic solver ls1 and visualized in situ with the visualization framework MegaMol.
[Picture: VISUS]
ISAV 2019: Tobias Rau presents his publication.
ISAV 2019: Tobias Rau presents his publication.

Contact

M. Sc.

Tobias Rau

Doctoral Researcher

Patrizia Ambrisi
M.A.

Patrizia Ambrisi

Public Relations SFB 1313

To the top of the page