How to configure namelist_quilt options for large domain

[jacob.stu.allen]

Hello!

I'm running WRF 4.6.0 over a 708x674 domain and looking for ways to make I/O more efficient. Is there some general wisdom on how nio_tasks_per_group or nio_groups should vary as a function of the amount of total nodes/processors or the domain size?

I'm currently using 14 N with 128 cpus each, but will likely scale down future runs to ~7-10 nodes.

Thanks in advance,
Jacob

 

[kwerner]

Hi Jacob,
Unfortunately we don't have a lot of detail to provide for the quilting option. We typically don't recommend using it unless you have a thorough understanding of computation, and we don't regularly test it. That being said, see this thread, on which a user has some recommendations on what worked for them. Maybe that will help!

 

[jacob.stu.allen]

Thank you very much @kwerner - that's just what I needed.