Cloud load testingAll features.
Everything a Cloud load test does today: the settings and their ranges, each figure it measures, how the rating is decided, and what is in the report.
29 features in 7 groups, checked against the product on 14 November 2026.
Setting up a run
One verified site
You choose the site from a list of your own verified sites. Every request is a GET of its home page over HTTPS.
50 to 1,000 virtual users
A slider in steps of 50; it starts at 200.
20 to 120 seconds
A slider in steps of 10; it starts at 30 seconds.
One level, held
The number of users you choose is held for the whole run, and each user sends one request after another without a pause. There is no ramp.
Run now or book for later
A single run can be booked for a time up to 14 days ahead, entered in your own time zone.
Sent from the cloud
The traffic comes from a dedicated machine in a cloud data centre in Virginia, USA, started for your run.
What it measures
Requests and failures
Total, successful and failed requests, and the error rate. Any answer that is not 2xx counts as a failure.
Requests per second
Over the whole run.
Response time
The average, the fastest, the median and the slowest, and the 90th, 95th and 99th percentiles.
Time to first byte
The average and the 95th and 99th percentiles.
Answers and data
The share of answers that were 2xx, and the total data received.
The rating and the summary
A risk rating by fixed rules
Elevated when 5% of requests or more failed or the 99th percentile reached 3 seconds. Moderate when 0.5% failed, the 95th percentile reached 1 second or the 99th reached 1.5 seconds. Otherwise Low. AI does not decide it.
Careful wording when requests fail
When 5% or more failed, the response times are labelled as including error answers, and are not called good.
An AI summary
Three short parts: the verdict, what it means and what to check, written from the run's figures; its verdict has to agree with the rating. The results are saved before the AI is asked, so a failed summary loses nothing.
The report
A PDF for every finished run
Cover, executive summary, risk scorecard, performance summary, scope and method, latency analysis, the raw figures, trend, and recommendations in order of priority.
Trend
The error rate and the 95th percentile beside the previous finished run on the same site. It says so when that run used a different number of users or a different length.
It states its limit
The report says that a formal pass or fail needs targets of your own.
Combined report
On the Reports page up to 50 reports, from this product and the others, can be merged into one PDF with a cover and a contents page.
History
The page shows your 10 most recent runs, with the PDF on each finished one.
Queue and controls
Queued, then started for you
A run joins a queue and starts by itself. The page tells you when the queue is busy.
One test at a time
A second cloud load test cannot be started while one of yours is queued or running.
Cancel from the Jobs page
A queued run can be cancelled, and a running one can be stopped.
A run that hangs is closed
A run with no result after 15 minutes is marked as failed.
Ownership and safety
Verified sites only
A run is accepted only for a site of yours whose DNS record has been confirmed, and the site is checked again when the run starts.
Public addresses only
A host that does not resolve to a public address is refused when you ask for the run and again before it starts.
Plans and limits
Business
50 cloud load tests a month. It is not part of Free or Pro.
Resets on the 1st
Allowances count over the calendar month, in UTC, and reset on the 1st.
A booked run counts when it runs
It is counted in the month it runs in, not the month you booked it.
Our failures are not counted
A run that never started, or failed on our side, does not use the allowance.
What it does not do.
- It holds one level of load. It does not climb through levels to find where the site gives way; Cloud stress testing shows how a site behaves as load climbs.
- It requests the home page only: no other paths, no sign-in and no journey through several pages.
- It records a failure as an answer that was not 2xx, with no count for each status code and no chart over time.
- It runs from one region, and from a different address each time, so there is no address to allow in advance.
- It is not a penetration test: nothing is exploited and your application's logic is not tested. We do not offer one.