From: Distributed systems

Distributed computer systems have evolved using message passing as
their main method of communication.  Other communication systems used
in loosely coupled distributed systems, such as RPC, are usually
implemented on top of an underlying message passing system.  On the
other hand, in tightly coupled systems, such as a multi-processor
machine, the communication method used is usually shared memory.

In distributed shared memory (DSM) systems [Nitzberg & Lo, 91],
processes share data transparently across node boundaries; data
faulting, location, and movement is handled by the underlying system.
Among other things, this allows parallel programs designed to use
shared memory to execute transparently on a loosely coupled
distributed system.  While the performance implications cannot be
ignored, the advantages of the shared memory programming model are
well known:

- Shared memory programs are usually shorter and easier to understand
  than equivalent message passing programs.

- Large or complex data structures may easily be communicated.

- Shared memory gives transparent process-to-process communication.

- Programming with shared memory is a well-understood problem.

Shared-memory (or `procedure-oriented') and message-oriented operating
systems are, in some sense, equivalent [Lauer & Needham, 78], though
it has been claimed that the former are `more powerful' [Tam et al.,
90].