Go C#

Limit Concurrency

In some scenarios, it can be useful to be able to limit the number of parallel interactions with some resource.

e.g. there might be some 3rd party service which cannot handle more than X simultaneous requests for whatever reasons, be it technical or legal.

In those cases, you can use a round-robin router to address this.

graph LR

l1(1)
class l1 message
l2(2)
class l2 message
l3(3)
class l3 message
l4(4)
class l4 message

r1(1)
class r1 message
r2(2)
class r2 message
r3(3)
class r3 message
r4(4)
class r4 message

lr((Limited Resource))

Router(Router)
class Router light-blue
Routee1(Worker1)
class Routee1 light-blue
Routee2(Worker2)
class Routee2 light-blue
Routee3(Worker3)
class Routee3 light-blue


l4---l3---l2---l1-->Router

Router---r4---r1-->Routee1
Router----r2-->Routee2
Router----r3-->Routee3

Routee1--->lr
Routee2--->lr
Routee3--->lr


As seen in the image above, we use a round-robin router to forward messages to, in this case, three worker actors. These worker actors can then interact with the limited resource.

Because actors process one message at a time, we can guarantee that there will never be more than (in this case) three parallel requests to the limited resource at any given point in time.

Limit Concurrency using Routers

«{ warning }» This content is dated, pre ActorSystem, PullRequests are welcome «{/ warning }>

In this post we will explore how we can use the actor model and routers in Proto.Actor to limit concurrency. That is, we want to ensure that no more than X concurrent workers are working at the same time. This can be useful when working with some form of expensive or limited resource, or when you want to scale up and maximize CPU core utilization.

First, let’s define a struct that will represent some form of work:

type workItem struct{ i int }

In this case, the work-item only carries an int, but this can of-course be replaced with something more meaningful.

Now let’s define our worker function, we do this by defining a function receiving an actor.Context object. The Context in this case is the context of your actor. that is, it contains the current message, children, and other elements of the actor.

func doWork(ctx actor.Context) {
	if msg, ok := ctx.Message().(*workItem); ok {
		//do work
	}
}

As seen above, we have to cast the current message into a *workItem pointer. Go lacks generics and an actor can handle many message types, this is why messages in Proto.Actor are of type interface{}. (Proto.Actor also work with typed “Grains”, more on that in another post)

Once we have received a message of the correct type, we can now perform some sort of work. The type of work you want to perform can be anything, in this example we will settle for a simple log statement:

func doWork(ctx actor.Context) {
	if msg, ok := ctx.Message().(*workItem); ok {
		log.Printf("%v got message %d", ctx.Self(), msg.i)
	}
}

Creating an actor

Now we need to create an actor and pass some work to this actor:

func main() {
	props := actor.FromFunc(doWork)
	pid := actor.Spawn(props)
	for i := 0; i < 1000; i++ {
		pid.Tell(&workItem{i})
	}
	console.ReadLine()
}

We start by calling actor.FromFunc to define how the actor should be created, in this case, we simply say that we want to create a stateless actor from a function. Then we call actor.Spawn to creates an instance of the actor, this is similar to the new function in Go. The actor.PID returned by Spawn is an identifier for the actor, and can be used to enqueue messages to the actors mailbox, We do this by calling pid.Tell(message).

Now we can send messages asynchronously to our actor, and the actor can process the messages one by one.

But what if we want to limit concurrency to any other number than 1?

One approach would be to create a slice and then spawn multiple actors from the same props, and then manually round-robin over those actors and send them each a message. This would result in x actors each pulling one message at a time from their own unique mailbox. There is however an easier way to do this. Proto.Actor has support for “Routers”. “Routers” are special actors that do exactly this.

Working with Routers

What we want in this specific case is a “Round-robin pool”, that is, we want a pool of actors, where the workload is sent in a round-robin manner to the workers. We can do this by altering the above code to this:

const maxConcurrency = 5

func main() {
	props := router.NewRoundRobinPool(maxConcurrency).WithFunc(doWork)
	pid := actor.Spawn(props)
	for i := 0; i < 1000; i++ {
		pid.Tell(&workItem{i})
	}
	console.ReadLine()
}

You can see how we replaced the previous actor.FromFunc with router.NewRoundRobinPool. This means that we now create a Props that will create a router with a round-robin strategy, and also spin up 5 workers, each using the doWork message handler. And this is it, now you can limit concurrency to a fixed number of workers, in our case 5. And whatever work you send to the router PID, is now guaranteed to only execute at a maximum concurrency level of 5.

There are of-course other strategies and use-cases for routers, but more on that in my next post.

Icon