could message passing be analogous to a microkernel design whereas shared address space is analogous to monolithic?

Do we observe any significant performance differences, as a general rule of thumb, between shared address space models and message passing models?

What are good ways of practically implementing message passing? How different is usage of messaging style communication from signal handling? Or is a lot of that abstracted away from the program? As best as I can tell, this is a good source to play around with MPI: https://www.open-mpi.org/ and get started.

@chii, I think the performance differences will depend on the size of the messages, and how much time it will take to send the message. I also assume if shared address space is possible, we will most likely to prefer it over message passing? Like in the photo, sending messages is the only way to communicate.