need to clarify one question first: what does the distributed hangman game mean or which part should be distributed? there are 2 types: stateless || stateful server stateless server might be the best way to solve the problem, since the game is very simple, nothing about security, and with a good encoding, client side can do all the caching of the game process with very cheap network communication. This makes the server very scalable.(in another sense, the server is just a skeleton to implement the logic. the game is just like running on users' machine and doing the calculation on server) stateful: this one is a little bit tricky. The server side may need to record users session obtained by users' cookie or the user ID. Considering millions of users playing at the same time, we can store the user session in a distributed database or file system like Chord File system or Google file system. using SHA-1 hash to distribute the user session, and do the calculation on that machine. To do the load balancing, we can use open source software like Haproxy as master node to dispatch the calculation. To implement some features like leader border: we might need to use a master node to implement a B+ tree or other better data structure (fix sized priority queue, if only first K records are needed)