WebPagetest Forums
Actual VS Page Tester - Printable Version

+- WebPagetest Forums (https://www.webpagetest.org/forums)
+-- Forum: WebPagetest (/forumdisplay.php?fid=7)
+--- Forum: Bugs/Issues (/forumdisplay.php?fid=10)
+--- Thread: Actual VS Page Tester (/showthread.php?tid=496)



Actual VS Page Tester - directive - 01-20-2011 09:46 PM

I have a client who is concerned about the speed that the page tester is showing (6.5-12 seconds).

However the site actually surfs to a human MUCH quicker.

Couple questions:

1. Does the page tester always test as if it's a completely fresh load? Like after clearing cache?

2. Does page tester simulate a certain internet speed?

I ask because we click through the site its going very fast, but the page tester is now kinda clouding the issue.


RE: Actual VS Page Tester - pmeenan - 01-21-2011 03:50 AM

Surfs to a human where, on what connectivity with what browser? :-D

1 - It tests both. The first view test loads with a completely clear cache and the repeat view is with a very recent cache (closes the browser, re-opens it and visits the same page again).

2 - By default it simulates a DSL connection with a 50ms last-mile ping time/RTT. You can configure it for whatever form of connectivity you'd like though. The testers are also physically located across the globe so the latency to the site depends on where the test was run from.

Usually we see situations where the site performs really fast on the developer's desktop because they are physically close to the server where the site is hosted on a really fast connection. That's actually the main reason the tool was originally built because we had developers on 100Mbit ethernet connections < 1ms from the data center where the sites were hosted so the pages always seemed fast, even though for real users with real latencies it was excruciatingly painful.

If you have a specific test you have questions about you can shoot the url to me (pmeenan@webpagetest.org) and I can take a look.

Thanks,

-Pat


RE: Actual VS Page Tester - directive - 01-21-2011 05:28 AM

(01-21-2011 03:50 AM)pmeenan Wrote:  Surfs to a human where, on what connectivity with what browser? :-D

1 - It tests both. The first view test loads with a completely clear cache and the repeat view is with a very recent cache (closes the browser, re-opens it and visits the same page again).

2 - By default it simulates a DSL connection with a 50ms last-mile ping time/RTT. You can configure it for whatever form of connectivity you'd like though. The testers are also physically located across the globe so the latency to the site depends on where the test was run from.

Usually we see situations where the site performs really fast on the developer's desktop because they are physically close to the server where the site is hosted on a really fast connection. That's actually the main reason the tool was originally built because we had developers on 100Mbit ethernet connections < 1ms from the data center where the sites were hosted so the pages always seemed fast, even though for real users with real latencies it was excruciatingly painful.

If you have a specific test you have questions about you can shoot the url to me (pmeenan@webpagetest.org) and I can take a look.

Thanks,

-Pat

Yeah for us its the opposite, we develop it from NY on their servers in FL. They are right on top of the server and think its slow.

The pages they are concerned about you have to be logged in to see, so we have been using the downloaded version of the aol page tester.

Wold you be interested in a little consulting work to tell me what you think? We'd need to hook up off the forums to get you access etc.


RE: Actual VS Page Tester - pmeenan - 01-21-2011 08:39 AM

The downloaded version just displays exactly what your browser loads, it doesn't do anything with the cache or network, that's only in the web-driven version.

Go ahead and hit me up offline and I can help out (not consulting per-say, I'm usually pretty good about carving out some time to help in general). You can actually script a login to a web site through the hosted version which would probably help get you a consistent reference point.

Thanks,

-Pat