Package org.terracotta.statistics.derived.latency

Class DefaultLatencyHistogramStatistic

java.lang.Object

org.terracotta.statistics.derived.latency.DefaultLatencyHistogramStatistic

All Implemented Interfaces:

LatencyHistogramQuery, LatencyHistogramStatistic, ChainedEventObserver, ChainedObserver

public class DefaultLatencyHistogramStatistic
extends java.lang.Object
implements LatencyHistogramStatistic, ChainedEventObserver

Field Summary

Fields

Modifier and Type	Field	Description
private BarSplittingBiasedHistogram	histogram
private long	nextPruning
private long	pruningDelay
private LatencyHistogramQuery	query
private java.util.function.LongSupplier	timeSupplier

Constructor Summary

Constructors

Constructor	Description
DefaultLatencyHistogramStatistic(double phi, int bucketCount, java.time.Duration window)
DefaultLatencyHistogramStatistic(double phi, int bucketCount, java.time.Duration window, java.util.function.LongSupplier timeSupplier)	Create a histogram maintained over a sliding time window.

Method Summary

Modifier and Type	Method	Description
java.util.List<Histogram.Bucket>	buckets()
long	count()
void	event(long time, long latency)	Called to indicate an event happened.
java.lang.Long	maximum()
java.lang.Long	minimum()
private static java.lang.Long	nullOrVal(double val)
java.lang.Long	percentile(double percent)
long[]	percentileBounds(double percent)
<T> T	query(java.util.function.Function<LatencyHistogramQuery,T> fn)	Enables to query the histogram several times within a synchronized state so that every calls are consistent
java.lang.String	toString()
private void	tryExpire(boolean force, java.util.function.LongSupplier time)

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.terracotta.statistics.derived.latency.LatencyHistogramQuery

median

Methods inherited from interface org.terracotta.statistics.derived.latency.LatencyHistogramStatistic

maximumStatistic, medianStatistic, minimumStatistic, percentileStatistic

Field Detail

histogram

private final BarSplittingBiasedHistogram histogram

timeSupplier

private final java.util.function.LongSupplier timeSupplier

pruningDelay

private final long pruningDelay

query

private final LatencyHistogramQuery query

nextPruning

private long nextPruning

Constructor Detail

DefaultLatencyHistogramStatistic

public DefaultLatencyHistogramStatistic(double phi,
                                        int bucketCount,
                                        java.time.Duration window,
                                        java.util.function.LongSupplier timeSupplier)

Create a histogram maintained over a sliding time window.

The constructed histogram is:

• maintained over window sliding window
• consists of bucketCount buckets
• where b1.size() ~= b0.size * phi

If "phi" is high, the quantile bounds will be more precise for lower percentiles such as minimum. If "phi" is low, the quantile bounds will be more precise for higher percentiles such as 99%-ile, maximum.

Parameters:

phi - histogram bucket bias factor

bucketCount - number of buckets

window - sliding window size, in ns

timeSupplier - the supplier of time, which must be in the same unit as the time passed to the {event(long, long)} method.

DefaultLatencyHistogramStatistic

public DefaultLatencyHistogramStatistic(double phi,
                                        int bucketCount,
                                        java.time.Duration window)

Method Detail

buckets

public java.util.List<Histogram.Bucket> buckets()

Specified by:

buckets in interface LatencyHistogramQuery

count

public long count()

Specified by:

count in interface LatencyHistogramQuery

minimum

public java.lang.Long minimum()

Specified by:

minimum in interface LatencyHistogramQuery

Returns:

The minimum value or null if no value

maximum

public java.lang.Long maximum()

Specified by:

maximum in interface LatencyHistogramQuery

Returns:

The maximum value or null if no value

percentile

public java.lang.Long percentile(double percent)

Specified by:

percentile in interface LatencyHistogramQuery

Parameters:

percent - the percentage (0.0-1.0)

Returns:

the value below which percent% of the observations may be found. Will return the upper bound of the approximated range.

percentileBounds

public long[] percentileBounds(double percent)

Specified by:

percentileBounds in interface LatencyHistogramQuery

Parameters:

percent - the percentage (0.0-1.0)

Returns:

the inclusive bounds (min and max) representing the range containing the right value for the given percent%-ile.

event

public void event(long time,
                  long latency)

Description copied from interface: ChainedEventObserver

Called to indicate an event happened.

Specified by:

event in interface ChainedEventObserver

Parameters:

time - the clock at event completion in ns

latency - the event latency in ns

query

public <T> T query(java.util.function.Function<LatencyHistogramQuery,T> fn)

Description copied from interface: LatencyHistogramStatistic

Enables to query the histogram several times within a synchronized state so that every calls are consistent

Specified by:

query in interface LatencyHistogramStatistic

toString

public java.lang.String toString()

Overrides:

toString in class java.lang.Object

tryExpire

private void tryExpire(boolean force,
                       java.util.function.LongSupplier time)

nullOrVal

private static java.lang.Long nullOrVal(double val)