The calculations within FAST have not yet been made parallel because there is not yet been a need to do so to handle the memory or computation speed requirements. Instead, multiple FAST simulations are often run in parallel (in batch mode).
That said, the modularization framework of FAST v8 was set up with future parallel processing in mind, and this may be a path to pursue in the future.
we are doing parallel computing by distributing thousands of single load cases (single FAST instances) into our network with several workstations using HTCondor. Condor is open source and manages the jobs within your cluster.
Each load case does not run so long. The problem is indeed more the quantity of load cases.
Alternatively microsoft HPC works as well (I used it with Flex5 in the past which is also one instance per DLC).
I would be interested to know if someone already tried to run multiple FAST Jobs on Amazon Cloud.
Have you ever figured out how many CPU cores or threads are suited when you are doing your simulation?
When I generate a big BTS file or simulate a big wind farm(by FAST.Farm), the OMP program on Linux system performs worse than the non-OMP program on Windows. My friends tell me more threads is not always better. So how do I choose the right thread?
That is surprising to me; I would generally expect the more threads, the faster FAST.Farm would exectute. I would suggest enabling compiler directive printthreads, recompiling, and reruning the simulation, which should provide some information about the thread usage when executing FAST.Farm.
You may also be interested in recent pull requests, such as the following, where we have introduced additional OpenMP parallelization in FAST.Farm and addition capability (such as simplified wind turbine model options in OpenFAST) aimed at enabling faster simulations of FAST.Farm for large wind farms:
Sorry for the repeated submission of the issue, as I am concerned that no one will reply to the latest message on Github, but it has now been resolved.
If anyone wants an answer here, you need to download the rc-3.5.1 branch of OpenFAST. I hope it can help you.