I feel like many things that do this don't make the trade offs clear enough, in the idea of making it more usable. But structuring your data across multiple pro

I feel like many things that do this don't make the trade offs clear enough, in the idea of making it more usable. But structuring your data across multiple processes is hard. Getting good performance in MPI can be very difficult, and even the topology of your processes across servers and the physical network can have an impact. I'm never convinced hiding this complexity is a good idea...