I am looking at setting up some website load testing scripts and need some help in finding a formula to estimate how many concurrent users are browsing a website at peak times, based on common metrics such as visits, average page views per visit, and average visit duration.
For example:
Peak visitors per hour: 1,000
Average page views per visitor: 3
Average time per visit: 5 minutes
Should I be considering at any other stats? Thanks in advance!
Yeah, because web traffic is very variable, an average of peak visitors over an hour may include most of them in the first half of the hour. Depending on how you calculated the peak, you may want to add a margin for that.
The easiest way is to calculate peak page requests/second and reverse engineer your visitors based on that. Also, peak requests may be initiated by some event, such as an email going out with a particular link or similar, the scenario in this case may be quite different from your normal traffic flows (e.g. single page visits).
It may pay to think about different test scenarios.
Another thing to think about in a content edited website is the effect of editors using the site during normal traffic, so adding an editor scenario to the test mix will give feedback on how these actions affect page responses.