Hi all,
didn’t find any results on this matter while searching this forum: I would like to run my simulations on a “low level” cluster by utilizing some workstations. Can I do this through the GUI by specifying the -hosts or a -hostfile only? And should any MFiX instance be “started up” on these hosts, or would the mpirun command initiate everything?
You can run parallel jobs from the GUI. Does your cluster use a queue like slurm or gridengine?
You will need to build a mpi enabled solver. You can do this through the GUI (make sure you have all the compiler dependencies loaded in the environment first):
Then when you went to run the model, press the play button, and select the solver you just built. If the solver is build with mpi, you’ll see the DMP options. You can also add additional flags if needed.
Thanks! I’m already running DPM locally. I was thinking if it is possible to “connect” to other hosts (computers really). It’s not a real cluster in that sense, I’m just trying to use leftover hardware we have here. Would I have to run this through slurm or gridengine? I was hoping that all I needed was to specify the hostfile with machines and cpu’s, and the -np. Question is, would this work if the other hosts (regular computers), are installed through anaconda? I’m not too confident on what is needed on the hosts end as you can see. Maybe all needed is the mpi, and the solver is provided by the main “node”?