Above the clouds: introducing Akka

Above the Clouds:
Introducing Akka
Jonas Bonér
CTO Typesafe
Twitter: @jboner

presented by Martin Odersky

The problem

It is way too hard to build:
1. correct highly concurrent systems
2. truly scalable systems
3. fault-tolerant systems that self-heals
...using “state-of-the-art” tools

Vision

Simpler
Concurrency
Scalability
Fault-tolerance

Vision

...with a single uniﬁed
Programming model
Runtime service

Replicate and distribute
for fault-tolerance

ARCHITECTURE

CORE
SERVICES

ARCHITECTURE

ADD-ON
MODULES

ARCHITECTURE

CLOUDY
AKKA

WHERE IS AKKA USED?
SOME EXAMPLES:

FINANCE TELECOM
• Stock trend Analysis & Simulation
• Streaming media network gateways

• Event-driven messaging systems
SIMULATION
BETTING & GAMING • 3D simulation engines

• Massive multiplayer online gaming
E-COMMERCE
• High throughput and transactional
• Social media community sites
betting

Event-
driven

Actor

Behavior

State

Akka Actors
one tool in the toolbox

Actors
case object Tick

class Counter extends Actor {
var counter = 0

def receive = {
case Tick =>
counter += 1
println(counter)
}
}

Create Actors

val counter = actorOf[Counter]

counter is an ActorRef

Start actors

val counter = actorOf[Counter].start

Stop actors

val counter = actorOf[Counter].start
counter.stop

Send: !

counter ! Tick

ﬁre-forget

Send: !!!
// returns a future
val future = actor !!! Message
future.await
val result = future.result

returns the Future directly

Future
val future1, future2, future3 =
Future.empty[String]

future1.await
future2.onComplete(f => ...)

future1.completeWithResult(...)
future2.completeWithException(...)
future3.completeWith(future2)

Future

val f1 = Futures.future(callable)

val f2 = Futures.ﬁrstCompletedOf(futures)
val f3 = Futures.reduce(futures)((x, y) => ..)
val f4 = Futures.fold(zero)(futures)((x, y) => ...)

Future
val future1 = for {
a: Int <- actor !!! "Hello" // returns 5
b: String <- actor !!! a // returns "10"
c: String <- actor !!! 7 // returns "14"
} yield b + "-" + c

val future2 = for {
a <- actor !!! Req("Hello") collect { case Res(x: Int) => x }
b <- actor !!! Req(a) collect { case Res(x: String) => x }
c <- actor !!! Req(7) collect { case Res(x: String) => x }
} yield b + "-" + c

Dataflow
import Future.flow

val x, y, z = Promise[Int]()

flow {
z << x() + y()
println("z = " + z())
}
flow { x << 40 }
flow { y << 2 }

Send: !!

val result = (actor !! Message).as[String]

uses Future under the hood and blocks until
timeout or completion

Reply
class SomeActor extends Actor {
def receive = {
case User(name) =>
// use reply
self.reply(“Hi ” + name)
}
}

HotSwap
self become {
// new body
case NewMessage =>
...
}

HotSwap
actor ! HotSwap {
// new body
case NewMessage =>
...
}

Set dispatcher
class MyActor extends Actor {
self.dispatcher = Dispatchers
.newThreadBasedDispatcher(self)

...
}

actor.dispatcher = dispatcher // before started

Remote Server
// use host & port in conﬁg
Actor.remote.start()

Actor.remote.start("localhost", 2552)

Scalable implementation based on
NIO (Netty) & Protobuf

Two types of
remote actors
Client initiated & managed
Server initiated & managed

Client-managed
supervision works across nodes

import Actor._

val service = remote.actorOf[MyActor](host, port)

service ! message

Server-managed
register and manage actor on server
client gets “dumb” proxy handle

import Actor._

remote.register(“service:id”, actorOf[MyService])

server part

Server-managed
val service = remote.actorFor(
“service:id”,
“darkstar”,
9999)

service ! message

client part

Server-managed
import Actor._

remote.register(actorOf[MyService])
remote.registerByUuid(actorOf[MyService])
remote.registerPerSession(
“service:id”, actorOf[MyService])

remote.unregister(“service:id”)
remote.unregister(actorRef)

server part

Remoting in Akka 1.1
Remote Actors
Client-managed
Server-managed

Problem
Deployment (local vs remote) is a dev decision
We get a ﬁxed and hard-coded topology
Can’t change it dynamically and adaptively

Needs to be a
deployment & runtime decision

Clustered Actors
(in development for upcoming Akka 2.0)

Address

val actor = actorOf[MyActor](“my-service”)

Bind the actor to a virtual address

Deployment
•Actor address is virtual and decoupled from how it
is deployed
•If no deployment conﬁguration exists then actor is
deployed as local
•The same system can be conﬁgured as distributed
without code change (even change at runtime)

•Write as local but deploy as distributed in the
cloud without code change
•Allows runtime to dynamically and adaptively
change topology

Deployment conﬁguration
akka {
actor {
deployment {
my-service {
router = "least-cpu"
clustered {
home = "node:test-node-1"
replicas = 3
stateless = on
}
}
}
}
}

akka { Address
actor {
deployment {
my-service {
clustered {
replicas = 3
stateless = on
}
}
}
}
}

akka { Address
actor { Type of load-
deployment {
my-service {
clustered {
replicas = 3
stateless = on
}
}
}
}
}

akka { Address
deployment {
my-service {
clustered {
Clustered replicas = 3
or Local stateless = on
}
}
}
}
}

akka { Address
deployment {
my-service {
clustered {
} Home node
}
}
}
}

akka { Address
deployment {
my-service {
clustered {
} Home node
}
}
} Nr of replicas
} in cluster

akka { Address
deployment {
my-service {
clustered {
} Home node
}
}
}
Stateful or Nr of replicas
} in cluster
Stateless

The runtime provides
•Subscription-based cluster membership service
•Highly available cluster registry for actors
•Automatic cluster-wide deployment
•Highly available centralized conﬁguration service
•Automatic replication with automatic fail-over upon
node crash
•Transparent and user-conﬁgurable load-balancing
•Transparent adaptive cluster rebalancing
•Leader election
•Durable mailboxes - guaranteed delivery

Upcoming features

•Publish/Subscribe (broker and broker-less)
•Compute Grid (MapReduce)
•Data Grid (querying etc)
•Distributed STM (Transactional Memory)
•Event Sourcing

Akka Node
val ping = actorOf[Ping](“ping”)
val pong = actorOf[Pong](“pong”)

ping ! Ball(pong)

Akka Node
val ping = actorOf[Ping](“ping”)
val pong = actorOf[Pong](“pong”)

ping ! Ball(pong)

Ping Pong

Akka Node

Akka
Cluster Node

Ping Pong

Akka
Cluster Node

Akka Node

Akka Akka
Cluster Node Cluster Node

Ping Pong

Akka Akka

Akka
Cluster Node

Akka Akka

Ping Pong

Akka Akka

Akka
Cluster Node

akka {
Akka actor { Akka
deployment {
Cluster Node ping {} Cluster Node
pong {
router = "round-robin"
clustered {
replicas = 3
stateless = on
}
}
Ping} Pong
}
}

Akka Akka

Akka
Cluster Node

akka {
Akka actor { Akka
Ping deployment {
pong {
clustered {
replicas = 3
stateless = on
}
}
} Pong
}
}

Akka Akka

Akka
Pong
Cluster Node

akka {
Akka actor { Akka
Ping deployment { Pong
pong {
clustered {
replicas = 3
stateless = on
}
}
}
}
}

Akka Akka
Pong

Akka
Pong
Cluster Node

akka {
Akka actor { Akka
Ping deployment { Pong
pong {
clustered {
ZooKeeper
ZooKeeper
replicas = 3
ZooKeeper
stateless = on
Ensemble
}
}
}
}
}

Akka Akka
Pong

...let’s take a
standard OO
application

Above the clouds: introducing Akka

Which components have
critically important state
and
explicit error handling?

Classiﬁcation of State
• Scratch data
• Static data
• Supplied at boot time
• Supplied by other components
• Dynamic data
• Data possible to recompute

• Input from other sources; data
that is impossible to recompute
•

Classiﬁcation of State
• Scratch data
• Static data
• Supplied Must time
at boot be
• Supplied by other components
protected
• Dynamic data
• Databy anyto recompute
possible means

• Input from other sources; data
that is impossible to recompute
•

Fault-tolerant
onion-layered
Error Kernel

Linking
link(actor)
unlink(actor)

startLink(actor)
spawnLink[MyActor]

Fault handlers
AllForOneStrategy(
errors,
maxNrOfRetries,
withinTimeRange)

OneForOneStrategy(
errors,
maxNrOfRetries,
withinTimeRange)

Supervision
class Supervisor extends Actor {
  faultHandler = AllForOneStrategy(
    List(classOf[IllegalStateException]) 
    5, 5000))

  def receive = {
    case Register(actor) => link(actor)
  }
}

Manage failure
class FaultTolerantService extends Actor {
...
override def preRestart(reason: Throwable) = {
... // clean up before restart
}
override def postRestart(reason: Throwable) = {
... // init after restart
}
}

...and much much more
STM
FSM
HTTP Camel
Microkernel Guice
OSGi
Dataﬂow
AMQP
scalaz Spring Security

Get it and learn more
http://akka.io

Above the clouds: introducing Akka

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Above the clouds: introducing Akka (20)

Recently uploaded (20)

Above the clouds: introducing Akka

Editor's Notes