Environment to simulate classic distributed computation model on single machine

Question

I'm searching for a tool to simulate the classic distributed computation model on a single machine to implement several algorithms in the purpose of the paper that I'm working on. Thus, performance isn't so important, it's only for scientific applications.

I want to have the possibility to specify the number of processes and the communication links between them. In other words, I want to define the network graph structure.

The computation should be asynchronous and message driven i.e. I want to send messages between connected processes and react to these messages.

Is there any library or framework for such computations? Simpler is better. Language doesn't matter so much, but I'd prefer Python or C++. I've given a look at Celery but I didn't find there the possibility to specify the connection between processes.

Answer 1

Very easy - python remote objects - pyro will run on a single machine or on multiple across a network.

Answer 2

You might also stick to python multiprocessing primitives, that allow you to create processes that communicate via queues, shared memory etc. Way easier than it seems ;)

See: http://docs.python.org/3/library/multiprocessing.html