[sersunzo]

Hi everybody,


is it possible that running a case in parallel affects the stability? I know it sound strange, but I tested exactly the same case with different processors and it looks like the higher the number of processors the smaller the Co number required in order to assure stability. I do not think it is a problem of how openFoam is installed in the cluster otherwise the simulation would not run at all.



It is a bit puzzling, did anybody experience something similar?



The test I am talking about is a Euler-Euler (solved with twoPhaseEulerFoam) gas-liquid 2D vertical channel 50mmx3mm (500x30 cells). The gas is injected from one of the two longer walls, top and bottom are inlet and outlets. I decomposed the channel for parallel calculation only along the longer dimension (y). I tried 1, 8 and 24 processors runs. The Co number in order to assure stability was respectively 1.0, 0.5, 0.1.



I know that some solvers does not scale very well with grids of less that 10,000 cells, but the issue I do not understand concerns the stability rather than the performance.


Thanks in advance


Alessio