The whole point of this system is to provide connectivity to the clients. The clients connect to a single communication server after credentials negotiation with the Coordinator Server.

The coordinator server is the key entry point of our communications system.

• It exposes a WS endpoint for the clients to negotiate the communication with the communications server
  • Caveat: It’s very important to notice the coordinator server will choose a communication server randomly, which means all communication servers should be equivalent to each other. That is, a client connected to a cluster, should have a consistent latency no matter which communication server he ends connecting to.
• It exposes a WS endpoint for the communication servers to
  • Negotiate connections with the clients
  • Discover other communications server in the cluster

The communications server is the heavy lifter of the system.

• It connects to the Coordinator Server via WS and uses this connection to discover other servers and negotiate client connections
• Every communication server may be connected to clients and
  • It relays packets from the clients to other clients
  • It relays packets to all the connected servers
  • It handles the business logic of the packets (topics, etc)
• It has to keep the WS connection alive, always. If the connection is closed, it has to retry until success.