search

radical-collaboration / hpc-workflows

NSF16514 EarthCube Project - Award Number:1639694
5 stars 0 forks source link

Use case description for PU simulations on Summit #61

Closed mturilli closed 4 years ago

mturilli commented 6 years ago

Create template document for this use case and share link here.

mturilli commented 6 years ago

This is a copy of the Y1 document. We can use it as a base to describe the changes required by Summit: https://docs.google.com/document/d/1uugJsRRSTHDhMFb3NWaRRPwyoMiNPKn6sr2iO3IqMaQ/edit#

wjlei1990 commented 5 years ago

There are a few things we planned to do on Summit. Just a quick note here for future discussion.

1) Increase the number of earthquakes. Currently, we have 1,840 earthquakes that are ready to be used . We are thinking about expanding the data to 4,000 or 6,000 in the future. The computational cost goes linearly with the number of earthquakes, if using the same simulation resolution.

2) Bring in more model parameters. Our parametrization is based upon elastic model with transversely isotropic assumption. We are planning to bring in Q inversion, which will turn the model into anelastic one. The computational cost will increase by ~60% if using the same resolution.

3) Increase the resolution. Our model resolution is 17 sec and on summit we are planning to push it to 9 sec. This will the most challenging part of our study. Double the resolution usually result in 10 times increase in computation time and storage.

uvaaland commented 5 years ago

The above bullets are accurate. In terms of priority, @jeroentromp says that it is 1, 3 and then 2. First increasing the number of events, then pushing the resolution to 9 sec, then considering additional model parameters.