How can a computer network be modulated and counted as a help?

Question

Programming

The task at first glance is simple, but googling did not lead to anything, if anyone has experience, please tell me.
Actually task:
1 to create a model of a computer network. (here I must say that I found nothing, only one program (omnet ++), but it is terrible and not convenient).
1.1 there is a server and N other computers (Clients)
in the network 2 The network must be programmed so that it can multiply two matrices. That is, the first Client multiplies the first row of the first matrix by the first column of the second matrix, the Second Client multiplies the second row of the first matrix by the second column of the second matrix, and so on. Then all Clients send their results to the Server which generates the final matrix.
The main task is to clearly show that a computer network (ala super computer) can perform certain operations (in our case, matrix multiplication).
PS: I will be grateful for any help. I am ready to clarify if something is not clear (for the first time I am writing a question on the net :)).

Answer 1

Well, as if the network itself has nothing to do with it. You need some library for network computing, for example MPI www.opennet.ru/docs/RUS/MPI_intro
habrahabr.ru/company/intel/blog/251357
habrahabr.ru/post/121235
or via http.

Answer 2

you can use MPI, but it needs a cluster to work. You can use self-made (there is one development at the ready), but calculations in such a network make sense only if their calculation takes more time (at least an order of magnitude) than transferring data for calculation over the network, that is, if you have a 100 * 100 matrix, it is more profitable parallelize on 1 machine than to drive data over the network for calculation. Roughly speaking, if your task "multiplying 1 row by 1 column" takes at least 10 ms to calculate, then it makes sense to think about the division of calculations, otherwise the overhead of data transmission in a computer network will eat up all the advantage in the distribution of calculations.

Answer 3

The technology is called Map-Reduce. Implemented, for example, Apache Hadoop, or Mongodb, but the task of mastering is quite difficult. Perhaps there are some other simpler options.