Performance enhancements and visualization for RSQSim earthquake simulator

Dmitry Pekurovsky, Amit Chourasia, Keith B. Richards-Dinger, Bruce E. Shaw, James H. Dieterich, & Yifeng Cui

Published August 15, 2016, SCEC Contribution #6933, 2016 SCEC Annual Meeting Poster #347

We report work in progress on RSQSim earthquake simulator code (Dieterich/Richards-Dinger). One goal is to streamline and optimize the algorithm to make it more scalable when running on high core counts on modern supercomputer platforms, such as Blue Waters (NCSA/UIUC), Stampede (TACC). Load imbalance is the major bottleneck inherent in the algorithm. It creates congestion and waste of resources by making some processors idly wait until others finish their calculation. To overcome load imbalance several algorithmic modifications are being considered. By changing the order of computations it is possible to reduce the amount of imbalanced compute load, and thus improve overall performance.

In addition, visualization work is underway.
RSQSim generates a seismic event catalog for several thousand years. This catalog includes rich detail about the quake which not only includes the magnitude, location and time, but also which patches got affected and the details associated. Effective study of this catalog requires visualization to provide context in space and time on where an event have occurred. To accomplish this goal we have created web browser based interactive visualization that shows a selected earthquake event on a map, and also indicated preceding and upcoming event trails as well as patches that were affected. This work involved creation of a database from raw files, this database can be easily queried to fetch necessary information on demand. The visualization routine, uses that database on the backend to fetch data on demand and plot it on the map, which is displayed on web browser. Additionally we display 1000 year time series visualization of earthquake events in the neighborhood of selected events. The interface enables the scientist to interactively explore the RSQSim seismic catalog.

Key Words
Earthquake simulations, RSQSim, performance optimization, scalability, visualization, load balancing, MPI

Citation
Pekurovsky, D., Chourasia, A., Richards-Dinger, K. B., Shaw, B. E., Dieterich, J. H., & Cui, Y. (2016, 08). Performance enhancements and visualization for RSQSim earthquake simulator. Poster Presentation at 2016 SCEC Annual Meeting.

Related Projects & Working Groups
Computational Science (CS)