I've thrown together a few different approaches to designing the multiplayer server, a few different
types of queues, and a few different threading models. The implimentations for the threading models
are in the backends folder and the implimentations for the queues are in the queues folder.
To compile and run the code, use python setup.py --queue <queue> --backend <backend>. To see the
list of backends and queues, run python setup.py --help. Try them out!
To run the server, use the python setup.py script to compile it and then run ./server. To run
a simulated client, run ./client. You can run as many clients as you like. Both the server and
client executables will log their activity to STDOUT. Both the client and the server take port and
host arguments via -p 8080 and -h 127.0.0.1 (which are the defaults), and the client takes an
additional -t 64 thread count argument which tells it how many clients to simulate.
To benchmark, try the different backends with different numbers of clients (by passing different thread counts to the client)! On my system, threadpool_backend handled up to around 64 concurrent clients, while at that level twothread_backend was falling behind. If a client doesn't get the same packet it send back within a 3 frame window, the script will tell you that, which reflects the server being overloaded. If the clients don't output anything, you know the server is serving packets within the 3 frame window!