i’m having a problem where at times the Repeat View (from cache) loads much slower than the First View.
to document that for the hoster i would need the Repeat View included in the compare page, to show that the Repeat View is not always delayed but still too often. they have a hard time to believe it, because when they do the test, there is no delay. in the compare view you could see the percentage of pages who have “Repeat View” problems.
hre’s an example:
Which “compare” page? You can include the repeat view in the filmstrip view for comparison but you need to manually build the URL (there are instructions on the filmstrip page). If you are referring to www.webpagetest.org/compare - that is meant to show just a very quick summary of what is possible with acceleration and isn’t supposed to be as in-depth as the normal WPT testing (and there is no reason a hoster should care about which page on WPT they are looking at).
In the normal UI you can do up to 10 runs, first and repeat view and see the results distribution: http://www.webpagetest.org/result/130506_AF_J76/ You can also click on the “plot full results” if you want to see what the distribution looks like: http://www.webpagetest.org/graph_page_data.php?test=130506_AF_J76&medianMetric=loadTime
At least in the test I ran, one of your first view runs ended up being the slow one so I think you have more of a variable performance problem than a specific issue with repeat views. To figure out what’s going on you’re going to have to instrument your back-end.
ok. really sophisticated. like it.
what i meant was, if i go to my own test history, check all the entries, and push ‘compare’ i get a comparison of my own history. and on that comparison page one of the tables is called “Timings”. but these are obviously only the ‘First View’ timings. what i would find helpful would be a second table ‘Timings’ which would show ‘Repeat View’ timings.
anyway. great tool, all in all.
Gotcha - the charts and tables are for the runs that are displayed in the filmstrip and the default comparison is to compare first views. Adding a -c:1 after any of the test ID’s would switch it to a cached view and you can add duplicate entries for first and repeat view. I’ll think about maybe an interstitial after you click the compare button asking if you’d also like to include repeat view tests (gets a bit confusing if you are looking at a lot of tests).
I’ve done this to our install and sent pmeenan a pullrequest.
Is a three line patch