Performance testing: load tests, percentiles, JMeter alternatives
Performance testing asks how a system behaves under load: how long a request takes at the 95th and 99th percentile, where it breaks, and what fails first. The two numbers that matter are the tail latency under a realistic mix of traffic and the point at which errors begin; an average hides both, which is why every load run here reports P50, P95, P99 and Apdex together.
Start with load-testing basics if the vocabulary is new, with intent-based load testing if you want a profile built from a recorded user journey rather than a hand-written script, and with the JMeter page if you are choosing a tool. The performance engine article explains what the numbers are measured from.
What people search for
The literal queries this hub is written to answer; each leads to one of the pages below.
- load testing tool
- performance testing tools
- p95 latency
- jmeter alternatives
- k6 vs jmeter
- stress testing tool
- load test from recorded session
Learn: the numbers
Percentiles, Apdex and virtual users, defined.
Load testing basics
Virtual users per plan, the percentiles you get, test data per virtual user, what the export carries.
P95 and P99 latency
Why the tail, not the mean.
Apdex
One score, and what it hides.
Virtual user
What one unit of load is.
Do it: build a load profile
From a recorded journey, not a script.
Intent-based load testing
A load profile from the flow users actually take.
Inside the performance engine
What is measured, from where, and the limits.
Load test
The definition, and the virtual-user caps per plan.
Peak-season readiness
What an online store tests before November.
Compare
The incumbents first.
JMeter alternatives
Who should stay on JMeter; no JMX import here, said plainly.
k6 export
A profile set up on the load screen leaves as a k6 script; the page says what the script carries and what it does not.
Load testing tool
The module, its caps and its limits.
Prove it: tools
Read a capture before you load-test it.
HAR file analyzer
The slowest requests and the server wait, from one page load.
Website analysis
Performance, accessibility and security from a single real page load.
Questions people ask
What is the difference between load testing and stress testing?
A load test applies the traffic you expect and checks the system meets its targets; a stress test raises the load past that until something fails, to learn what breaks first and whether it recovers. The same profile serves both; the ramp and the stopping rule differ. The load-testing basics guide covers both shapes.
Why report P95 and P99 instead of the average response time?
Because the average hides the slow requests that users remember. If 5% of requests take four seconds and the rest take 200 milliseconds, the mean is near 400 milliseconds and looks fine; P95 is four seconds and tells the truth. Every run here reports P50, P95, P99 and P99.9.
Can I export a load test to run it myself?
Yes, as a k6 script. On the load screen, Export as k6 script saves a script with the targets, virtual users, ramp-up, duration and wait of the profile; an attached data set is saved as data.csv with a second button, and the same section saves the profile as a JSON file the screen reads back. A plan built from a recording, with its steps, assertions and thresholds, has no export; the script carries a profile set up by hand on the load screen. There is no JMeter, Locust or Gatling export and no JMX import; the JMeter alternatives page lists what does transfer when a team moves.
Related topics
Try it on your own application
The Free plan asks for no card and runs for 14 days after the trial: record a flow, replay it, and read the evidence each step produced.