Skip to content
Tutorials Center
My learning

Topics

Lesson 18 of 37 ยท Assertions and Listeners

JMeter Listeners and Results: Tree, Summary and Aggregate Reports

Choose the right JMeter listener: View Results Tree, Summary Report, Aggregate Report, Response Time Graph and Simple Data Writer, and learn the columns, percentiles and costs of each.

Listeners show or record what happened. They are the windows into your test, and also its most common source of self-inflicted damage: a listener that keeps every response body in memory can crash JMeter long before the server shows any stress. This lesson explains the main listeners, how to read their columns and when to use each.

Honesty note. JMeter cannot run here. Listener names, columns and settings are real. The runnable blocks are stand-ins that compute the same statistics (average, percentiles, error percent, throughput) from sample data, so you can see exactly how the numbers are derived.

Add listeners with Add > Listener. Scope applies: a listener under the Test Plan records all samples; under a controller or sampler, only those below it. Several listeners can coexist, but each one adds cost.

View Results Tree

Shows each sample with tabs Sampler result, Request and Response data (with renderers such as Text, HTML, JSON, XML and Regexp Tester). Green means success, red failure. It is the best debugging tool and the worst load-test tool: it stores request and response data for every sample.

Rules:

  • Use it with 1 to 5 threads while building the script.
  • Tick Log/Display Only: Errors if you do use it in a bigger run, so only failures are kept.
  • Disable or delete it before real load tests. In non-GUI mode, listeners that draw in the GUI do not render anyway, but they still consume memory if present, so remove them.

Summary Report

A compact table with one row per sample label and a TOTAL row:

Column Meaning
# Samples Number of samples
Average / Min / Max Response time in ms
Std. Dev. Spread of response times
Error % Share of failed samples
Throughput Samples per second (or minute/hour, auto-scaled)
Received / Sent KB/sec Bandwidth
Avg. Bytes Mean response size

Light on memory because it keeps only totals, not each sample. Good for a quick look.

Aggregate Report

Like the Summary Report, but adds Median and the percentile columns 90% Line, 95% Line and 99% Line. The "90% Line" is the time within which 90% of samples completed. Percentiles tell you what slower users experience; the Average alone does not. It keeps all sample times in memory to compute percentiles, so it is heavier than the Summary Report. There is also Aggregate Graph, which charts the same data, and a Response Times Percentiles graph from plugins.

Graph and plugin listeners

  • Graph Results: legacy chart of samples, average, median and deviation. Fine for a quick view.
  • Response Time Graph: average per time interval for chosen labels.
  • Active Threads Over Time, Hits per Second, Transactions per Second, Response Times Over Time, Response Codes per Second (the "jpgc" plugins via Plugins Manager): the charts you will most want, but in a load test generate them after the run from the results file rather than live.
  • Backend Listener: streams metrics to InfluxDB or Graphite for Grafana dashboards during the run (built in, needs a backend).
  • Simple Data Writer: writes results to a file with no display and almost no overhead. Use it for real runs (or just -l results.jtl on the command line).

Result files from the command line

In non-GUI mode, -l results.jtl already records every sample, so you rarely need any listener inside the plan for load runs. Open the .jtl file afterwards in a listener (the Browse... field of any listener can load a results file), use the -e -o report-folder HTML dashboard, or analyze it with your own scripts. See JTL files and saving results.

The data behind the numbers: a worked example

The stand-in computes a Summary/Aggregate style table from raw samples, using nearest-rank percentiles as many reports do. JMeter's own percentile calculation can differ slightly in rounding (it uses its own interpolation rules), so treat this as an illustration of the idea.

JS Stand-in: compute an Aggregate Report from samples

Press Run (or Ctrl + Enter) to see the console output here.

No output.

Look at the Checkout row: with three samples the 90th and 95th percentiles equal the maximum, which is why small sample counts make percentiles meaningless. Report percentiles only with plenty of samples.

Listener costs

JS Stand-in: estimate listener memory

Press Run (or Ctrl + Enter) to see the console output here.

No output.

These are rough illustrations, not JMeter measurements, but they show why the Results Tree must never be left on during load.

Choosing listeners: a recipe

Situation Listeners
Building and debugging a script View Results Tree (1 to 5 threads), Debug Sampler
Quick check with moderate load in the GUI Summary Report
Reading percentiles by label Aggregate Report
Real load test (non-GUI) None in the plan, or Simple Data Writer; use -l and the HTML dashboard
Live dashboards Backend Listener with InfluxDB and Grafana

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. Which listener is best for inspecting a single request and its response body?

    The Results Tree shows request and response details for each sample.
  2. 2. Which report adds the 90%, 95% and 99% lines?

    Aggregate Report extends the summary with the median and percentile columns.
  3. 3. Why remove View Results Tree before a big load test?

    Keeping all sample data can exhaust JMeter's heap and distort results.
  4. 4. What is the lightest way to record results for a non-GUI run?

    -l writes one line per sample with minimal overhead; analyze it later.

Summary

  • Listeners show or save results; scope follows the tree and each one costs resources.
  • View Results Tree is for debugging with few threads; Summary and Aggregate Reports show statistics and percentiles.
  • Percentiles need many samples to be meaningful.
  • For real load tests record with -l or a Simple Data Writer and analyze afterwards.
  • Backend Listener sends live metrics to InfluxDB/Graphite for Grafana.

Next: JTL Files and Saving Results