In situ run becomes first of its kind to exceed one million Message Passing Interface processes.
Kitware has scheduled two times, during which it will deliver a presentation that provides exclusive insight on the largest-known in situ simulation run at its booth (3437) at The International Conference for High Performance Computing, Networking, Storage and Analysis (SC16). The presentation times will begin on Monday, November 14, 2016, at 7:30 p.m. MDT and on Tuesday, November 15, 2016, at 11:00 a.m. MDT, respectively.
Kitware participated on a team that worked with collaborators to complete the run for a project earlier this year. As part of the project, the team aims to create an infrastructure that can execute algorithms that users write across in situ software implementations. The particular implementation that the team used for the largest-known in situ simulation run combined ParaView Catalyst and SENSEI.
“ParaView Catalyst provides a lightweight ParaView server library, and SENSEI offers an interface that transfers analysis code among in situ infrastructures,” said Andrew Bauer, Ph.D., the lead developer of ParaView Catalyst at Kitware. “ParaView Catalyst and SENSEI streamline the simulation pipeline. Together, they enabled us to complete a larger run than ever before.”
For the run, ParaView Catalyst and SENSEI computed the Parallel Hierarchic Adaptive Stabilized Transient Analysis (PHASTA) simulation code. The run employed a 6.3 billion-cell unstructured grid and used 32,768 out of 49,152 nodes on Mira, an IBM Blue Gene/Q supercomputer from Argonne Leadership Computing Facility at Argonne National Laboratory. With 1,048,576 Message Passing Interface (MPI) processes on over 500,000 cores, the run quadrupled the size of the previous largest-known in situ simulation run. While the previous run also computed PHASTA with ParaView Catalyst, it did not utilize SENSEI.
“With ParaView Catalyst, users can specify which visualization and analysis capabilities of ParaView they seek to implement,” Bauer said. “Now, with SENSEI, users can better leverage in situ infrastructures to reduce runtime.”
The total runtime for the simulation clocked in at 653 seconds. Of this total, the time for in situ computation comprised 13 percent. During SC16, Kitware will address how information gleaned from the simulation run guided the development community to update ParaView. With these updates, the team projects that the total runtime for the same simulation would now come to 601 seconds, and the time for in situ computation would reduce to 5.6 percent of the total.
The SC16 presentation will complement a paper that team members wrote with collaborators for the conference. The paper, “Performance Analysis, Design Considerations, and Applications of Extreme-scale In Situ Infrastructures,” reviews in situ computation and shares additional results of the largest-known in situ simulation run. As well as Bauer, authors on the paper include Utkarsh Ayachit, Earl P. N. Duque, Greg Eisenhauer, Nicola Ferrier, Junmin Gu, Kenneth E. Jansen, Burlen Loring, Zarija Lukić, Suresh Menon, Dmitriy Morozov, Patrick O’Leary, Reetesh Ranjan, Michel Rasquin, Christopher P. Stone, Venkat Vishwanath, Gunther H. Weber, Brad Whitlock, Matthew Wolf, K. John Wu and E. Wes Bethel. Bethel will discuss the paper at the State-of-the-Practice: System Characterization and Design session at the conference. The session will take place on Thursday, November 17, 2016, from 4:00 p.m. to 4:30 p.m. MDT in room 355-D of the Salt Palace Convention Center.
To learn more about the largest-known in situ simulation run, SC16 attendees can visit the Kitware booth during the conference exhibition from November 14 to November 17, 2016. Along with the exhibition, Kitware will participate in a number of conference activities that highlight its technology and expertise in high-performance computing and visualization. For details on scheduled activities, please refer to the event listing on the Kitware blog. To set up a time to meet with Kitware team members at SC16, please email kitware(at)kitware.com.
This material is based upon work supported by the U.S. Department of Energy, Office of Science, under Award Number DE-SC0012387.
This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.