There are centralized address servers that list the addresses of those decentralized data serves.

So the client connects to one of the main centralized servers and receives the latest list of data servers, and then the client can connect to those decentralized data servers to look for files.