Question 1

What is website load testing?

Accepted Answer

Website load testing sends many concurrent virtual users through the pages of a website or web application and measures response time, throughput and error rate as the load grows. It tells you how many visitors the site serves before pages slow down or requests fail. For planning a test and reading its results, see the load testing guide.

Question 2

How do I load test a website?

Accepted Answer

Create a JMeter project, run a plan from it and open the job. In the HyperExecute Projects dashboard, click New Project, upload your test files and save. Select the .jmx, click Run, fill in the run form and click Run Test. Results appear under Jobs. If there is no plan yet, the HTTP(S) Test Script Recorder in Apache JMeter builds one from a browser session, and the JMeter load test setup documentation lists every field.

Question 3

Is a website load test the same as a website speed test?

Accepted Answer

No. A speed test loads one page for one visitor and reports how long that page took. It says nothing about the thousandth visitor arriving at the same moment. A page that scores well in a speed test can still fail under load, because a single request never fills a connection pool or a database queue. The web performance testing guide covers both.

Question 4

Does TestMu AI load test with real browsers?

Accepted Answer

No. Driving a real browser for every virtual user, for example through a WebDriver Sampler, makes each one consume the CPU and memory of a full browser session. Our JMeter documentation notes that this can cut the users one machine sustains by an order of magnitude, so load is generated at the HTTP level and browser checks are left to Selenium or Playwright runs.

Question 5

How many concurrent users can I simulate on my website?

Accepted Answer

From 100 on Free to 8,000 on Pro per license, with a custom ceiling on Enterprise. How close a run gets depends on the work each virtual user does. The median desktop page makes 77 requests according to the 2025 HTTP Archive Web Almanac, so a virtual user that fetches page assets, with Retrieve All Embedded Resources switched on in JMeter, costs far more than one that requests only the HTML. Large payloads and long chained journeys lower the number further.

Question 6

What is a virtual user hour?

Accepted Answer

A virtual user hour is one virtual user running for one hour, and it is the unit every plan is metered in. A run of 2,000 users for 30 minutes uses 1,000 of them. Higher user counts and longer tests both draw the monthly allowance down faster, so a soak test that holds load for hours uses far more than a short spike at the same user count.

Question 7

What does per license mean in the plan limits?

Accepted Answer

Each license carries its own limits, and licenses stack. The pricing page gives the example of two Pro licenses providing up to 40 load generators. Pro is the top self-serve tier. Beyond stacked licenses, Enterprise is custom priced for higher concurrency, longer test durations and dedicated load generators.

Question 8

Why did my run generate more users than the plan specified?

Accepted Answer

Without portal overrides, the thread count in the .jmx is repeated on every machine in every region. A 250 user plan on 3 machines in 2 regions generates 1,500 virtual users, and uses six times the virtual user hours. Enter Total Users on the run form to have the platform split that number across regions and machines. We flag this before a first run because it changes both the result and the bill.

Question 9

Will a WAF or rate limiter interfere with a load test?

Accepted Answer

Yes, unless the load generators are allowlisted there too. A WAF, bot filter or rate limiter sees thousands of requests arriving from a handful of addresses and starts throttling or blocking them, so the run measures the filter instead of the site. Add HyperExecute's static outbound IPs to its allowlist for the test window. Only test a site you own or have written permission to load.

Question 10

Which JMeter settings matter most for a website plan?

Accepted Answer

Timeouts matter most. Set explicit Connect Timeout and Response Timeout values on HTTP Request Defaults, so a slow request fails and releases its thread instead of piling up. We also recommend a Duration Assertion for the same reason. On very large runs, setting jmeter.save.saveservice.subresults to false keeps the results file small.

Question 11

Can I use Gatling or k6 instead of JMeter?

Accepted Answer

Yes for Gatling, with caveats for k6. A Gatling project in the portal offers named Capacity, Stress and Soak test types. k6 is documented on HyperExecute as a framework integration, driven by YAML and the CLI, with announced version support capped at 0.52. For .jmx plans, see JMeter load testing.

Question 12

Can I automate website load tests in my CI pipeline?

Accepted Answer

Yes, with Gatling. A Gatling run is configured in a hyperexecute.yaml committed next to the simulation, so the pipeline triggers it through the HyperExecute CLI and the load shape is reviewable in a pull request. Its Gatling HTML report comes back as a job artifact. JMeter runs are configured in the Projects portal and have no CLI or YAML path. Gatling load testing covers both paths.

Question 13

Is website load testing free on TestMu AI?

Accepted Answer

Yes. The Free plan costs $0 and needs no credit card. Both paid HyperExecute plans, MultiOS and Linux Only, also include 100 virtual user hours per month per license. The performance testing pricing page lists every tier.

Question 14

Do I also need to load test my API?

Accepted Answer

Usually, yes. A website plan reaches the API only through the pages that call it. An API load testing run requests each endpoint directly, so its throughput and error rate are measured on their own. The plan formats and load generators are the same for both, and the cloud load testing tool overview lists the limits they share.

Question 15

How do I read a load test that failed?

Accepted Answer

By the pattern of the failures. Response times that climb and never recover suggest a thread or connection leak in the plan. Failures confined to the ramp-up mean the ramp was too steep for the site or the machine allocation. Failures from one region only point to the load distribution setting or the network path to that region.

Question 16

Can I export or filter load test results?

Accepted Answer

Yes. Request Stats and the Errors report each have a Download as CSV button. The Timeline Report filters by region, virtual users, hits, errors, response time, latency, connect time and response code. Request Stats also lists 90th, 95th and 99th percentile response times beside the average, minimum and maximum. Metrics can be switched between milliseconds and seconds.

Question 17

What else can I set on the run form?

Accepted Answer

Job Labels and the Java version. Labels tag a run so it can be traced in logs and reports, for example by build number. Plans run on Java 11 by default, and Java 25 can be selected instead.

Website Load Testing for Sites and Web Apps

Inside a Website Load Test

Turn One Real Visit Into the Test

Rehearse the Rush Before It Arrives

Send Traffic From Where Visitors Are

Find the Page That Gives Out First

Test the Site Before It Goes Public

See the Page the Way Visitors Do

Reproduce Website Traffic, Request by Request

Website Load Testing Plans and Limits

Some Love from our Customers

More Reasons to LoveTestMu AI (formerly LambdaTest)

TestMu AI Named a Challenger in the 2025 Gartner® Magic Quadrant™

TestMu AI recognized in The Forrester Wave™: Autonomous Testing Platforms, Q4 2025

Wall of Fame

Enterprise-Grade Security

Integrations

Frequently Asked Questions