Hi

I am running the parallel version of fluent on a cluster of 8 nodes. I want to evaluate the performance (speedup). Is there anybody having suitable case and data files and the speedup they obtained on their system with those files? I want to compare the results.

Thanks

hi,

it varies from case to case and the optimum no of nodes for best performance depends on the communication between nodes for your case. but its still easy. u can measure it urself. run once on a stand alone m/c and note down time taken for, say 50 timesteps. then repeat it for different no of nodes. remember that number of nodes need not mean a linear boost-up always!

muppur

I have used this method, and I have obtained the speedup graphs, but I need to compare it with someone else's works. I need somebody that has the same experience, so I can use the same case and compare the results (speedup) with his/her work.