Distributed deployment of Message Classifier

[Buddhima]

In this first phase we have worked towards a well functioning solution. For that we have used distributed-worker (master-worker) example of Typesafe Activator. But at the last stages we recognized that it permits actual distributed deployment.
It will be great to modify this project, in such a way that this can be deploy in a real-world distributed nodes.

[shakya]

Normal distributed akka projects use separate JVMs and/or separate physical machines for the nodes in the cluster. System administrator has to start those nodes manually at the beginning. But distributed-worker (master-worker) example has simplified this aspect by running all nodes in the same JVM. We realized this at the latter stages of the project.

At that stage in order to support separate JVMs for separate actor systems we had to change the entire project structure. Since this was not in our main project scope and we were lacking the time we decided to keep the project structure as it was and not to deploy it on separate nodes.

Akka also supports the concept called "remote actors" where master can deploy actors on remote nodes without manually setting them up. For that we have to use the actor provider "akka.remote.RemoteActorRefProvider". But the example which we built on uses the actor provider "akka.cluster.ClusterActorRefProvider". "akka.remote.RemoteActorRefProvider" cannot be used with "akka.cluster.ClusterActorRefProvider" because the latter internally uses the prior. Therefore that approach too was not feasible.

references
https://groups.google.com/forum/#!topic/akka-user/9tonp_5LOWE
http://doc.akka.io/docs/akka/snapshot/java/remoting.html
http://doc.akka.io/docs/akka/snapshot/java/cluster-usage.html