CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   HD Space for LES (http://www.cfd-online.com/Forums/main/3125-hd-space-les.html)

ANINDYA February 13, 2001 12:55

HD Space for LES
 
I would like to know the approximate disk space required to run a 3-D LES problem. My dimensions are 100 X 100 X 50. Please let me know how I can determine the size of the Hard disk space required for running different dimensions. I believe that LES produces more output data than other turbulence models(1 and 2 equations).

Thanks a lot

Anindya

Jens Chr. Bennetsen February 14, 2001 04:25

Re: HD Space for LES
 
Hi,

Well well, The amount of disk space required to store data from LES is easy to determine.

Let say you want to store all time steps from a simulation for post-processing asssumning 8 byte data representation and NX, NY, NZ the dimensions, we get:

NX x NY` x NZ x 8 byte x No. of Time Step

In this way you are able to do post-processing of all point in your domain.

But your could also only store data for each time step at pre-choosen point that you are particular interested in. In this way you do not need large Disk drives for data. Also remenber, that I/O are costly in terms of cpu time !

Regards

Jens

Madhukar M. Rao February 15, 2001 13:48

Re: HD Space for LES
 
Hello,

You also need to consider multiple dependent variables: typically 3 velocity components + Pressure ?

"NX x NY` x NZ x 8 byte x No. of Time Step" x No. of variables.

If he is using Fortran unformatted files, then most Fortran compilers add record markers at the start and end of each record, adding a finite overhead.

If he is saving as text "ascii" files, then he is going to use more than 8 bytes per number, possibly double the space.

Cheers,

Madhu


andy February 16, 2001 07:58

Re: HD Space for LES
 
I think the implication of the previous two replies may be somewhat misleading. It is not normal to store:

100*100*50*100000*4*4 = 800 GBytes

of information for a modest LES run (assuming 100000 time steps, 4 variables (u,v,w,p) and single precision values).

LES runs are typically performed a bit more like experiments. You decide what you want beforehand and the LES run assembles it as it progresses. Typically for would assemble:

(i) Mean values everywhere

(ii) Mean Reynolds stresses everywhere

(iii) Possibly various sub-grid stresses if this is of interest.

(iv) p.d.f.s at a few interesting locations.

(v) Anything else that is normally obtained by exeriment.

(vi) Of course, time evolutions on a few planes to make the obligatory movies. Probably leaving out a few grid points and almost certainly skipping quite a few time steps.

If you miss something then you run the simulation again. It is rare to get everything right the first time particularly if you are new to LES.


Zhengtong Xie February 16, 2001 12:56

Re: HD Space for LES
 
It is not neccessary to store all of the time steps such as the first part steps.

It would be better the code has the "RESTART" ability. Then you could rerun the code from the former case's last step.


All times are GMT -4. The time now is 05:45.