com.lazerycode.jmeter.analyzer.statistics.Samples Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of jmeter-analysis-maven-plugin Show documentation
Show all versions of jmeter-analysis-maven-plugin Show documentation
Parses JMeter result files and computes performance indicators such as average request duration
package com.lazerycode.jmeter.analyzer.statistics;
import java.util.*;
/**
* Collects samples (as a sequence of values of type "long") and provides the following values:
*
*
* - min: Minimum value
* - max: Maximum value
* - average: Average value
* - standard deviation
* - samples per second
* - quantiles
*
*
* @author Dennis Homann, Arne Franken, Peter Kaul
*/
public class Samples {
private static final float SECOND = 1000f;
// number or error samples
private long errors = 0;
// number of success samples
private long success = 0;
// collected samples
private List samples = new ArrayList();
// timestamps corresponding to samples
private List timestamps = new ArrayList();
// minimum sample timestamp
private long minTimestamp = Long.MAX_VALUE;
// maximum sample timestamp
private long maxTimestamp = Long.MIN_VALUE;
// minimum sample value
private long min = Long.MAX_VALUE;
// maximum sample value
private long max = Long.MIN_VALUE;
private volatile boolean finished = false;
// current number of samples which are aggregated into a single sample
private int compression = 1;
// maximum number of samples to store
private final int maxSamplesCount;
private List samplesBuffer = new ArrayList();
private List timestampsBuffer = new ArrayList();
private double total = 0;
// sum of all values each powered by 2
private double totalPowered2 = 0;
private long standardDeviation;
// The value histogram
private Map histogram;
// ----------------------
/**
* Creates a new instance where a limited number of samples is stored internally.
* If more samples are added, existing samples will be compressed internally.
*
* @param maxSamples The maximum number of samples. 0=store no samples
* @param histogram If set to true the a value histogram should be counted as well.
*/
public Samples(int maxSamples, boolean histogram) {
this.maxSamplesCount = maxSamples;
if( histogram ) {
this.histogram = new HashMap();
}
}
/**
* Adds an "error" sample. An error sample isn't used for statistics values such as average, ...
*
* @param timestamp The timestamp of the sample. It's assumed that a timestamp is greater or equal than the previous one
*/
public void addError(long timestamp) {
assertNotFinished();
errors++;
setTimestamp(timestamp);
}
/**
* Adds a "success" sample
*
* @param timestamp The timestamp of the sample. It's assumed that a timestamp is greater or equal than the previous one
* @param value The sample value, e.g. the response duration or response bytes
*
* @see #addError(long)
*/
public void addSample(long timestamp, long value) {
assertNotFinished();
// handle counters/statistics
success++;
total += value;
totalPowered2 += Math.pow(value, 2);
//set min / max value
if( value > max ) {
max = value;
}
if( value < min ) {
min = value;
}
//set min / max timestamp
setTimestamp(timestamp);
//collect the value
if( histogram != null ) {
ValueCount count = histogram.get(value);
if( count == null ) {
//there was no other request with the same response value, collect
count = new ValueCount(value);
histogram.put(value, count);
}
//there already was another request with the same response value, increment
count.increment();
}
// store sample
add(timestamp, value);
}
/**
* Marks collecting samples as "finished"
*/
public void finish() {
finished = true;
// flush buffer by adding remaining items
if( samplesBuffer.size() > 0 ) {
addAggregated(samplesBuffer, timestampsBuffer);
samplesBuffer.clear();
timestampsBuffer.clear();
}
// adjust some statistics
// Standard Deviation: http://en.wikipedia.org/wiki/Standard_deviation#Rapid_calculation_methods
double totalPowered0 = success;
double totalPowered1 = total;
standardDeviation = (long) (Math.sqrt(totalPowered0 * totalPowered2 - Math.pow(totalPowered1, 2)) / totalPowered0);
// protect collected data against modification
samples = Collections.unmodifiableList(samples);
timestamps = Collections.unmodifiableList(timestamps);
}
/**
* @return Determines whether a relevant number of samples have been provided so that statistics can be computed
*/
public boolean hasSamples() {
return getSuccessCount() > 0;
}
public List getSamples() {
assertFinished();
return samples;
}
/**
* @return The sample's timestamp.
*/
public List getTimestamps() {
assertFinished();
return timestamps;
}
/**
* @return Total number of stored samples
* @see #getSuccessCount()
*/
public long getStoredSamplesCount() {
assertFinished();
return samples.size();
}
/**
* @return Total number of successful samples
*/
public long getSuccessCount() {
assertFinished();
return success;
}
/**
* @return Total number of samples having an error
*/
public long getErrorsCount() {
assertFinished();
return errors;
}
public long getMin() {
assertFinished();
if( !hasSamples() ) {
throw new IllegalStateException("No samples");
}
return min;
}
public long getMax() {
assertFinished();
if( !hasSamples() ) {
throw new IllegalStateException("No samples");
}
return max;
}
/**
* @return The average for all samples
*/
public long getAverage() {
assertFinished();
long count = getSuccessCount();
if( count == 0 ) {
throw new IllegalStateException("No samples");
}
return (long) total / count;
}
public long getTotal() {
assertFinished();
return (long)total;
}
public long getStandardDeviation() {
return standardDeviation;
}
public long getMaxTimestamp() {
assertFinished();
return maxTimestamp;
}
public long getMinTimestamp() {
assertFinished();
return minTimestamp;
}
/**
* @return Number of successful samples per second
*/
public long getSuccessPerSecond() {
assertFinished();
long duration = getDuration();
if( duration == 0 ) {
return 0; // shouldn't happen
}
return getSuccessCount() / duration;
}
/**
* @return The duration in s
*/
public long getDuration() {
assertFinished();
return Math.round((getMaxTimestamp()-getMinTimestamp()) / SECOND);
}
/**
* Returns a Quantile with the grade/resolution q using counts as values
*
* @param q the grade
*
* @return the q-quantile.
*/
public Quantile getQuantiles(int q) {
assertFinished();
if( histogram == null ) {
throw new IllegalStateException("No histogram available");
}
return new Quantile(q, histogram.values());
}
//====================================================================================================================
private void assertNotFinished() {
if( finished ) {
throw new IllegalStateException("Already finished");
}
}
private void assertFinished() {
if( !finished ) {
throw new IllegalStateException("Not finished");
}
}
/**
* set min / max timestamp
* @param timestamp the timestamp
*/
private void setTimestamp(long timestamp) {
if( timestamp < minTimestamp ) {
minTimestamp = timestamp;
}
if( timestamp > maxTimestamp ) {
maxTimestamp = timestamp;
}
}
/**
* Collect timestamp and value
*/
private void add(long timestamp, long value) {
if( maxSamplesCount == 0 ) {
return;
}
// Disabling the compression if maxSamplesCount < 0
if (maxSamplesCount > 0) {
// check whether the maximum of samples is reached and reduce number of samples if necessary
if( samples.size() >= maxSamplesCount ) {
// compress
halve();
compression *= 2;
}
}
// add current sample
if( compression == 1 ) {
// store samples
samples.add(value);
timestamps.add(timestamp);
}
else {
// buffer samples for aggregation
samplesBuffer.add(value);
timestampsBuffer.add(timestamp);
if( samplesBuffer.size() >= compression ) {
// we have collected enough items
addAggregated(samplesBuffer, timestampsBuffer);
samplesBuffer.clear();
timestampsBuffer.clear();
}
}
}
/**
* Aggregates samples and timestamps and add them as a single item to samples/timestamp
*
* @param samplesBuffer samples to be aggregated. will be cleared
* @param timestampsBuffer timestamps to be aggregated
*/
private void addAggregated(List samplesBuffer, List timestampsBuffer) {
long firstTimestamp = timestampsBuffer.get(0);
long lastTimestamp = timestampsBuffer.get(timestampsBuffer.size()-1);
long aggregatedTimestamp = firstTimestamp + (lastTimestamp-firstTimestamp) / 2;
long aggregatedSample = 0;
for( long sample : samplesBuffer ) {
aggregatedSample += sample;
}
aggregatedSample = aggregatedSample / samplesBuffer.size();
samples.add(aggregatedSample);
timestamps.add(aggregatedTimestamp);
}
/**
* Cuts a list of samples in half by aggregating pairs a samples
*/
private void halve() {
List newSamples = new ArrayList();
List newTimestamps = new ArrayList();
Iterator si = samples.iterator();
Iterator ti = timestamps.iterator();
while( si.hasNext() ) {
long sample = si.next();
long timestamp = ti.next();
if( !si.hasNext() ) {
// there is no second sample. thus, don't aggregate this last element
newSamples.add(sample);
newTimestamps.add(timestamp);
}
else {
long secondTimestamp = ti.next();
long secondSample = si.next();
long aggregatedSample = (sample+secondSample) / 2;
long aggregatedTimestamp = (timestamp+(secondTimestamp-timestamp) / 2);
newSamples.add(aggregatedSample);
newTimestamps.add(aggregatedTimestamp);
}
}
samples = newSamples;
timestamps = newTimestamps;
}
}