Internal Structure
Like every other network, the hydra is build from different interconnected nodes. A node can either be a dedicated or a virtual machine. It is even possible to connected two JVMs on the same host machine.
The network is fully interconnected, allowing every node to directly communicate with the others. By monitoring the TCP connections, the nodes can easily detect, if a node is still up and reachable.
The nodes share a global registry, which holds information about a network structure. Each time a new nodes connects into the network, it will receive a eplica of the registry. It is seen as consensus, that the registry is always in sync. The ability to force a re-sync will be added later.
The network shares an eventstream, acting as an abstraction layer for the interprocess communication.
A single Node inside the net has two main layers. On the bottom layer, their is an actor system, encapsulating every process, so they behave like being atomic.
Each actor can listen to specific messages inside the eventstream and can send messages in response. The system will also generate eventmessages, if for example a new node connects to the network.
In the layer above the actor system, their are services. A service can be everything, from a single (low level) database query to a complex business application, build from other lower level services.
This layer holds the "userland" - the application forming code. As a result, it should by atypical to have the need to write an actor as part of a user service or application.
