I have been looking around on Google for sites pertaining to statistics of HTTP requests made during individual user sessions. I've come up empty. What I am attampting to do is to simulate the behaviour of individual users thinking for various lengths of time. I know Astra Load Test allows for think time. We are using a tool in which think time is customizable for individual HTTP requests themselves.
Does anyone know of any web sites that keep statistics of times between requests made in executing normal flow for things like posting information to forums or ordering products?
I've not seen anything like this, although I'd be interested to see it if it does exist. I don't know many people who even record this information or how you would get it from the logs. I guess things like webtrends could tell you, but then it's unlikely people will publish details of their webtrends reports.
The obvious places to look (I guess you've already looked but just incase...)
1) Existing logs. If you're upgrading from an existing service you might be able to analyse your current logs. The Microsoft tool LogParser allows you to run SQL queries against log files, which will help with the analysis.
2) Business Analysis - do they have any ideas of how long the process should take, is this something you could push back on them for?
3) User Experience - get a bunch of people off the street (or use friends/relatives) and ask them to go through the process, use that as a guide.
It might be easier to run a test test, one with a short thinktime, one with a large think time and see how it changes your results, make wild assumptions based on this test and press on.
From my experience of load testing, it's based on guesswork and speculation. Even though you get very specific numbers out at the end (response time = 1.345sec), you're making so many assumptions at the start that it might not be worth worrying about think time.
I'd be interested if other more experienced people disagree on this last point.
Everywhere's within walking distance if you have enough time.
Think times per request might be a tough route to go. To a user, clicking a button to perform a function may, to them, seem like one request but in reality spawns 15-20 http requests to various locations. So to a user, it may take 3 seconds, but in reality, it's 15 requests having response time of a few hundred milliseconds.