WebPagetest Forums
Questions regarding WebPagetest browser instances - Printable Version

+- WebPagetest Forums (https://www.webpagetest.org/forums)
+-- Forum: Web Performance (/forumdisplay.php?fid=3)
+--- Forum: Testing Tools/Services (/forumdisplay.php?fid=6)
+--- Thread: Questions regarding WebPagetest browser instances (/showthread.php?tid=13003)



Questions regarding WebPagetest browser instances - bananabeatz - 05-31-2014 06:34 AM

I apologize if these questions have been answered elsewhere, I'm a webpagetest novice Confused I tried looking through the forums and the documentation in search of answers but came up short.

I'm trying to wrap my head around how the browser instances are set up at the different webpagetest locations.

Are tests run on real browsers and devices, or are they simulated instances running on the servers?

How can I identify which browser version is used for the different instances (e.g. Chrome, Safari, Firefox, Chrome on Nexus 5, etc.)?

All responses appreciated Smile


RE: Questions regarding WebPagetest browser instances - pmeenan - 05-31-2014 07:24 AM

The tests are on real browsers and devices. The guaranteed failsafe way to check the version is to look at the User Agent string on a request but generally Chrome and Firefox are the latest stable versions (they auto-update) and Safari is the last build they did for Windows which was 5.something.

When a new Chrome version is being released there may be a day or two where the versions on different agents may not match until the roll-out is complete but otherwise they track whatever the current version is.


RE: Questions regarding WebPagetest browser instances - bananabeatz - 06-02-2014 11:10 PM

Okay, thanks so much for the feedback. For some reason I didn't think of inspecting the UA-string :-S

A small follow-up question: Are requests on shaped 3G-connection routed through an actual cell tower?


RE: Questions regarding WebPagetest browser instances - pmeenan - 06-02-2014 11:30 PM

No, they are on WiFi routed through a BSD bridge running dummynet. We've found that you get a LOT more consistent results staying off of the actual carrier networks which is good for perf testing (though it does mean the actual user experience will vary quite wildly).


RE: Questions regarding WebPagetest browser instances - bananabeatz - 06-03-2014 03:23 AM

Allright, good to know Smile Thanks again.


RE: Questions regarding WebPagetest browser instances - sidbala - 06-03-2014 10:30 AM

(06-02-2014 11:30 PM)pmeenan Wrote:  No, they are on WiFi routed through a BSD bridge running dummynet. We've found that you get a LOT more consistent results staying off of the actual carrier networks which is good for perf testing (though it does mean the actual user experience will vary quite wildly).

Patrick, is there any info on the end-to-end configuration of the BSD bridge? Is this simply a computer with two network interfaces - one connected to the WiFi router and another to the wired LAN that contains the host for Webpagereplay?


RE: Questions regarding WebPagetest browser instances - pmeenan - 06-05-2014 12:30 AM

There's no webpagereplay involved (all of the testing tests live sites) but otherwise yes. More specifically:


Phone ---(wifi)--- Apple Airport Access Point --- BSD Bridge --- Egress Router ->

There are a few other switches in the mix but the BSD bridge is just a low-end server (Atom supermicro) with 2 interfaces running freebsd (We've found that BSD tends to work a lot better for dummynet compared to linux).


RE: Questions regarding WebPagetest browser instances - pmeenan - 06-05-2014 12:32 AM

Oh, and I use Apple's access points exclusively because:

1 - They work. Never had to reboot one which is WAY more than I can say for any of the consumer routers from anybody else
2 - They support running as access points so I can just bridge them to my network

The test phones use their own SSID that runs on a frequency that is sufficiently separated from any others in use to avoid interference.