On the SDR Colombia datastack, I noticed that it runs more slowly on the slurm node than it does on my laptop, which is surprising. This might be part of the explanation.
In this htop screenshot, it appears that the 4 worker processes are all sharing core 1, each getting about 25% utilization. I'd expect each process to run on a different core and get close to 100% utilization each. For comparison, on my laptop at the same point in the model, each of the 4 processes is using >100% CPU:
On the SDR Colombia datastack, I noticed that it runs more slowly on the slurm node than it does on my laptop, which is surprising. This might be part of the explanation.
In this
htopscreenshot, it appears that the 4 worker processes are all sharing core 1, each getting about 25% utilization. I'd expect each process to run on a different core and get close to 100% utilization each. For comparison, on my laptop at the same point in the model, each of the 4 processes is using >100% CPU: