Scheduled Downtime
On Friday 21 April 2023 @ 5pm MT, this website will be down for maintenance and expected to return online the morning of 24 April 2023 at the latest

Timing issues

This post was from a previous version of the WRF&MPAS-A Support Forum. New replies have been disabled and if you have follow up questions related to this post, then please start a new thread from the forum home page.

antonleontiev

New member
Hello! I have installed the WRF model to the new computer (with Cent OS 7), which is much more fast than my previous. My aim is to obtain wrfout files every 3 hours. As I understand, I should create met_em files using WPS and then change value of history_interval at namelist.input to 180, which means that the output to wrfout files will be written every 3 hours, which were created successfully. But there is a problem with timing. The dx and dy parameters of my domain are 3000, so I need to set value of time_step to 18. It leads to the situation, when I see at the output something like this:
WRF TILE 1 IS 1 IE 150 JS 1 JE 300
WRF NUMBER OF TILES = 1
Timing for main: time 2018-01-01_00:00:18 on domain 1: 7.21617 elapsed seconds
Timing for main: time 2018-01-01_00:00:36 on domain 1: 3.12540 elapsed seconds
Timing for main: time 2018-01-01_00:00:54 on domain 1: 3.13478 elapsed seconds
Timing for main: time 2018-01-01_00:01:12 on domain 1: 3.13609 elapsed seconds
Timing for main: time 2018-01-01_00:01:30 on domain 1: 3.13895 elapsed seconds
So, I am in the situation, when calculation of data every 3 hours will take about 30-40 minutes. But at the computer, which I used before, these calculations took less time, than now. My namelist.input as attached, could anyone, please, tell me, what am I doing wrong? I don't need any calculations every 18 seconds, I need only every 3 hours.
 
Hi,
The history_interval is simply the time interval for files to be written out. The time_step is how often calculations are done. So in every history file, there should be many time steps. Unfortunately, in order to maintain stability in the model, you must keep the time step small (no more than about 6xDX). Can you attach your full namelist.input file and let me know how many processors you are using? I can check to make sure that seems reasonable.

Kelly
 
kwerner said:
Hi,
The history_interval is simply the time interval for files to be written out. The time_step is how often calculations are done. So in every history file, there should be many time steps. Unfortunately, in order to maintain stability in the model, you must keep the time step small (no more than about 6xDX). Can you attach your full namelist.input file and let me know how many processors you are using? I can check to make sure that seems reasonable.

Kelly
Thank you very much for your reply! You can find namelist.input at the attachment
 

Attachments

  • namelist.input
    3.1 KB · Views: 80
Thanks. Can you also let me know how many processors you are using to run this? It may be useful to attach your wrf out file (e.g., rsl.out.0000).

Kelly
 
Hello! I use 32 processors, but I don't see a significant time difference in time in comparison with 1 processor. Sorry, I can not send you an rsl file, because it does not exist (I don't know why).
 
Hi,
Thanks for sending that. Your domain is not huge, but it's big enough that, with the high resolution, running on too few processors is slowing down your simulation. If you have access to more processors, you should try to use more. I'm not sure how many cores you have on each node, but try something between 60 and 80 - depending on what works out best with your machine set-up).
 
Top