cynthiaz92 / oacp Goto Github PK

Observable Atomic Consistency Protocol is a memory model for distribued system. We extend the current GSP with monotonic update and try to get better performance. It's a combination between total order broadcast (which will be implemented through raft) and Conflict Free Data Type(CRDT). So our implementation will be divided into two stages:

1. Raft protocol
2. CRDT representation (start with GCounter, ORset)

Extend existing trait CRDT:

case class RGCounter(...) extends CRDT[Array[Int], Int]

object CRDT{

  implicit def RGCounterCRDT = new RGCounter
  
  ...

}

trait RGCounterProtocol {
  sealed trait CounterCmnd
  case class Add(num: Int) extends CounterCmnd
  case object Get extends CounterCmnd
  case class ResultIs(mValue: Int) extends CounterCmnd
}

class CounterClient extends OACPClient[Array[Int], Int, String] {

  val CounterClientBehavior: Receive = {
  
    ...
    
  }

  override def receive = CounterClientBehavior.orElse(super.receive)
  
  ...
  
}

eventuate-crdt0.10: (https://github.com/RBMHTechnology/eventuate/tree/r-0.10)

The code is included in src/multi-jvm/scala/se/kth/csc/progsys/oacp/ExampleSpec.scala This example uses twitter server and client and includes:

1. create three server actors(Raft cluster) and one client actor
2. send message to the raft cluster (msg type: AddFollower(CvOp), Twitter(TOp), Read(TOp))
3. count the number of messages between client and servers [Optional]

Running command: using "sbt" to the interaction mode, then "multi-jvm:testOnly se.kth.csc.progsys.oacp.ExampleSpec".