org.apache.kafka.server.util

Class ThroughputThrottler

java.lang.Object

org.apache.kafka.server.util.ThroughputThrottler

public class ThroughputThrottler
extends Object

This class helps producers throttle throughput. If targetThroughput >= 0, the resulting average throughput will be approximately min(targetThroughput, maximumPossibleThroughput). If targetThroughput < 0, no throttling will occur. To use, do this between successive send attempts:

     
      if (throttler.shouldThrottle(...)) {
          throttler.throttle();
      }
     
 

Note that this can be used to throttle message throughput or data throughput.

Constructor Summary

Constructors

Constructor and Description
ThroughputThrottler(long targetThroughput, long startMs)

Method Summary

Modifier and Type	Method and Description
boolean	shouldThrottle(long amountSoFar, long sendStartMs)
void	throttle() Occasionally blocks for small amounts of time to achieve targetThroughput.
void	wakeup() Wakeup the throttler if its sleeping.

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

ThroughputThrottler

public ThroughputThrottler(long targetThroughput,
                           long startMs)

Parameters:

targetThroughput - Can be messages/sec or bytes/sec

startMs - When the very first message is sent

Method Detail

shouldThrottle

public boolean shouldThrottle(long amountSoFar,
                              long sendStartMs)

Parameters:

amountSoFar - bytes produced so far if you want to throttle data throughput, or messages produced so far if you want to throttle message throughput.

sendStartMs - timestamp of the most recently sent message

Returns:

throttle

public void throttle()

Occasionally blocks for small amounts of time to achieve targetThroughput. Note that if targetThroughput is 0, this will block extremely aggressively.

wakeup

public void wakeup()

Wakeup the throttler if its sleeping.