Question

我正在尝试测试Akka Streams的吞吐量，并了解随着请求数量的增加它的扩展程度。

我目前面临的问题是stream并发工作。我的stream由flows组成，每个人都会睡一秒钟以模拟功能。会发生的是，对于通过流传递的每个元素，流程将同步处理它。我希望这也是异步的，以优化我的表现。

这是我使用的代码：

// Flow that's being used
def processingStage(name: String): Flow[TestObject, TestObject, NotUsed] =
    Flow[TestObject].map { s ⇒
      println(name + " started processing " + s + " on thread " + Thread.currentThread().getName)
      Thread.sleep(1000) // Simulate long processing *don't sleep in your real code!*
      println(name + " finished processing " + s)
      s
    }

// Stream
def startStream() = {
        val completion = Source[TestObject](list.toList)
          .via(processingStage("A")).async
          .via(processingStage("B")).async
          .runWith(Sink.foreach(s ⇒ println("Got output " + s)))
  }

Answer 1

不要使用Thread.sleep（1000）来模拟延迟 - 使用基于时间的组合器。此外，如果要在同一个流中强制执行多个阶段的并发，请使用.async分界。有关详细信息，请参阅文档。

Answer 2

Streams默认是串行的。如果您希望并行处理流中的元素，则必须直接请求它。

文档描述了实现此目标的一种方法：http://doc.akka.io/docs/akka/2.4/scala/stream/stream-cookbook.html#Balancing_jobs_to_a_fixed_pool_of_workers

这是它添加到您的代码中：

  func addViewToNav()
  {
    myView = UIView(frame: CGRect(x: 0, y: 20, width: (self.navigationController?.navigationBar.frame.size.width)!-40, height: (self.navigationController?.navigationBar.frame.size.height)!))

    imageView.frame = CGRect(x: 5, y: (myView?.frame.size.height)!/2-14, width: 28, height: 28)
    let image = UIImage(named: "download")
    imageView.contentMode=UIViewContentMode.scaleAspectFill
    imageView.image=image
    imageView.layer.borderColor = UIColor.white.cgColor
    imageView.layer.cornerRadius = imageView.frame.size.width/2
    imageView.clipsToBounds = true
    imageView.backgroundColor=UIColor.yellow
    label = UILabel(frame: CGRect(x: imageView.frame.size.width+15, y: 0, width: 150, height:  (myView?.frame.size.height)!))
    label?.text="Label"
    label?.textColor=UIColor.white
    myView?.addSubview(imageView)
    myView?.addSubview(label!)
    self.navigationController?.view.addSubview(myView!)
    //add tap gesturte on imageview & label
    let tapGesture = UITapGestureRecognizer(target: self, action: #selector(self.tapBlurButton(_:)))
    myView?.isUserInteractionEnabled=true
    tapGesture.numberOfTapsRequired=1
    myView?.addGestureRecognizer(tapGesture)
  }

Answer 3

以下是您的代码的更新版本：

import akka.stream.scaladsl.{Sink, Source}

import scala.concurrent.Future 
import scala.concurrent.ExecutionContext.Implicits.global

object Test extends App {    

  // Flow that's being used
  def processingStage(name: String): Future[String] = Future {
    println(name + " started processing " + name + " on thread " + Thread.currentThread().getName)
    Thread.sleep(1000) // Simulate long processing *don't sleep in your real code!*
    println(name + " finished processing " + name)
    name
  }

  // Stream
  def startStream() = {
    val parallelism = 10 //max number of parallel instances
    val list = (1 to 1000000).toList.map(_.toString) //sample input 

    Source[String](list)
      .mapAsync(parallelism)(processingStage) //handles the items concurrently 
      .runWith(Sink.foreach(s ⇒ println("Got output " + s)))
  }

  startStream()
}

第一点是，您应该将processingStage函数转换为返回Future的方法。通过这样做，您可以更好地模拟并发任务。
其次，您应该使用mapAsync方法分阶段支持并发。据我了解，这是您正在寻找的确切功能。

Akka流并发流

3 个答案: