Investigate load-balancing and parallel efficiency

[manodeep]

Run scaling tests to find optimal allocation, grouping and partitioning of MPI tasks to the separate components.

Needs to account for NUMA architecture on the targeted queues

[JhanSrbinovsky]

if in the area can you profile the model so we can demonstrate how long it spends in each section. ~12 years ago, profiling suggested that the coupled model spent < 5% of the time at the land-surface.

[manodeep]

@JhanSrbinovsky That should be possible. I am early in the info gathering process and that will inform the design - so feedback like this is timely and welcome :)