Large output files for 3D airfoil simulation

[Bharathwaj Dharani]

Hello All,

I'm currently running a 3D airfoil simulation with approximately 170,000 elements, and my output "0.f00000*" field files are around 980 MB each.

Previously, for a 2D airfoil simulation with 17,000 elements, the corresponding field files were about 15 MB.

Is it typical for 3D simulation output files to be this much larger, or are there best practices I can implement to reduce their size?

My goal is to start running the 3d case,  gradually increase the Reynolds number (Re) for these simulations, starting with lower Re cases.

Any advice or insights would be greatly appreciated!

Thanks,

Bharathwaj

[YuHsiang Lan]

Hi Bharathwaj,

The size of a typical checkpoint file is

   E * lx1^ldim * Nfld * (4 bytes per float) + extra tables for elements map and min/max.

where 

  Nfld = 3 mesh + 3 velocity + 1 pressure = 7, or

  Nfld = 3 velocity + 1 pressure = 4

For example,

  2D, E=17,000, lx1=6, Nfld=7:   15.9 MB (decimal)

  3D, E=170,000, lx1=6, Nfld=7: 957.5MB (decimal)

Basically, you get 10x elements and another 6x from lx1.

60x file size is reasonable.

One thing you can do is to only keep the last two of the checkpoint files for the restart (1 for backup in case simulation dies in writing).

See full_restart, which dumps 2 sets of 3 files.

If the mesh is not moving, you can get rid of mesh via ifxyo=.false., so Nfld is three less.

As long as the very first file has mesh, which should be the default in Nek5000, you can still open it in paraview / visit.

You can also not save the pressure field (ifpo=.false.), which saves one of Nfld.

Hope this helps,

Yu-Hsiang

--