Résumé | In distributed computing, there are 2 main communication abstractions
considered : message-passing (processes directly send messages to one
another), and shared memory (processes communicate through some shared
object).
Under some conditions, it is possible for processes using a
message-passing system to implement (or simulate) some shared memory
objects. This is useful because it is often easier to write distributed
algorithms using those shared objects. Thus, any problem that can be
solved using shared memory, could also be solved in message-passing systems.
If some conditions are not verified in the message-passing system, it is
impossible to simulate usual shared memory objects (with good
properties). My work is about the simulation of some new (extended)
shared objects, with weaker properties, but that can still be
implemented under those circumstances.
|