Theater is an actor system that allows you to built scalable applications that are distributed by nature and run massively parallel.
In an actor system we take message passing between objects very literal. Instead of invoking method on objects the communication between objects only happens by sending it a message.
In theater you can define your objects as actors. Each actor can send and receive messages. Simply sending a message to an actor that does not exist will create one. Each actor is allowed to have state, just like real objects. Theater runs by default in a clustered configuratino with a gossip network for discovery of nodes. Actors are distributed acros the cluster for scalability. Add more nodes and you can add more actors. By storing the state in a shared repository we can easily migrate actors form VM to VM, from node to node.
Theater can be used for massive parallel computation. For example:
- We make one actor that reads a file with input data. Each line of input get send to one of the ten processing actors.
- Each processing actor does some form of processing on the line of data. The result is send to the result actor.
- The result actor aggregates all the results and e-mails them to the initiator.
With this setup we can scale the computation from a single machine with 1 core to a cluster of 10 machines. Also, we can scale the cluster while the computation is running.