com.netflix.concurrency.limits.limit

Class GradientLimit.Builder

java.lang.Object

com.netflix.concurrency.limits.limit.GradientLimit.Builder

Enclosing class:

GradientLimit

public static class GradientLimit.Builder
extends java.lang.Object

Constructor Summary

Constructors

Constructor and Description
Builder()

Method Summary

Modifier and Type	Method and Description
GradientLimit.Builder	backoffRatio(double backoffRatio) Ratio applied to the limit when a timeout was identified within the sampling window.
GradientLimit	build()
GradientLimit.Builder	initialLimit(int initialLimit) Initial limit used by the limiter
GradientLimit.Builder	maxConcurrency(int maxConcurrency) Maximum allowable concurrency.
GradientLimit.Builder	metricRegistry(MetricRegistry registry) Registry for reporting metrics about the limiter's internal state.
GradientLimit.Builder	minLimit(int minLimit) Minimum concurrency limit allowed.
GradientLimit.Builder	minRttThreshold(long minRttThreshold, java.util.concurrent.TimeUnit units) Deprecated.
GradientLimit.Builder	probeInterval(int probeInterval) The limiter will probe for a new noload RTT every probeInterval updates.
GradientLimit.Builder	probeMultiplier(int probeMultiplier) Deprecated.
GradientLimit.Builder	queueSize(java.util.function.Function<java.lang.Integer,java.lang.Integer> queueSize) Function to dynamically determine the amount the estimated limit can grow while latencies remain low as a function of the current limit.
GradientLimit.Builder	queueSize(int queueSize) Fixed amount the estimated limit can grow while latencies remain low
GradientLimit.Builder	rttTolerance(double rttTolerance) Tolerance for changes in minimum latency.
GradientLimit.Builder	smoothing(double smoothing) Smoothing factor to limit how aggressively the estimated limit can shrink when queuing has been detected.

Methods inherited from class java.lang.Object

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

Constructor Detail

Builder

public Builder()

Method Detail

minRttThreshold

@Deprecated
public GradientLimit.Builder minRttThreshold(long minRttThreshold,
                                                         java.util.concurrent.TimeUnit units)

Deprecated.

Minimum threshold for accepting a new rtt sample. Any RTT lower than this threshold will be discarded.

Parameters:

minRttThreshold -

units -

Returns:

Chainable builder

initialLimit

public GradientLimit.Builder initialLimit(int initialLimit)

Initial limit used by the limiter

Parameters:

initialLimit -

Returns:

Chainable builder

minLimit

public GradientLimit.Builder minLimit(int minLimit)

Minimum concurrency limit allowed. The minimum helps prevent the algorithm from adjust the limit too far down. Note that this limit is not desirable when use as backpressure for batch apps.

Parameters:

minLimit -

Returns:

Chainable builder

rttTolerance

public GradientLimit.Builder rttTolerance(double rttTolerance)

Tolerance for changes in minimum latency.

Parameters:

rttTolerance - Value >= 1.0 indicating how much change in minimum latency is acceptable before reducing the limit. For example, a value of 2.0 means that a 2x increase in latency is acceptable.

Returns:

Chainable builder

maxConcurrency

public GradientLimit.Builder maxConcurrency(int maxConcurrency)

Maximum allowable concurrency. Any estimated concurrency will be capped at this value

Parameters:

maxConcurrency -

Returns:

Chainable builder

queueSize

public GradientLimit.Builder queueSize(int queueSize)

Fixed amount the estimated limit can grow while latencies remain low

Parameters:

queueSize -

Returns:

Chainable builder

queueSize

public GradientLimit.Builder queueSize(java.util.function.Function<java.lang.Integer,java.lang.Integer> queueSize)

Function to dynamically determine the amount the estimated limit can grow while latencies remain low as a function of the current limit.

Parameters:

queueSize -

Returns:

Chainable builder

smoothing

public GradientLimit.Builder smoothing(double smoothing)

Smoothing factor to limit how aggressively the estimated limit can shrink when queuing has been detected.

Parameters:

smoothing - Value of 0.0 to 1.0 where 1.0 means the limit is completely replicated by the new estimate.

Returns:

Chainable builder

metricRegistry

public GradientLimit.Builder metricRegistry(MetricRegistry registry)

Registry for reporting metrics about the limiter's internal state.

Parameters:

registry -

Returns:

Chainable builder

backoffRatio

public GradientLimit.Builder backoffRatio(double backoffRatio)

Ratio applied to the limit when a timeout was identified within the sampling window. The default value is 0.9. A value of 1.0 means no backoff.

Parameters:

backoffRatio -

Returns:

probeMultiplier

@Deprecated
public GradientLimit.Builder probeMultiplier(int probeMultiplier)

Deprecated.

probeInterval

public GradientLimit.Builder probeInterval(int probeInterval)

The limiter will probe for a new noload RTT every probeInterval updates. Default value is 1000. Set to -1 to disable

Parameters:

probeInterval -

Returns:

Chainable builder

build

public GradientLimit build()