A couple of weeks ago I added a test where we hit a couple of pages as a logged in user. That server is located in NYC in Digital Oceans data center and the tests looks like this:
https://en.wikipedia.org/wiki/Main_Page -> https://en.wikipedia.org/wiki/Barack_Obama -> https://en.wikipedia.org/wiki/Democratic_Party_(United_States)
vs login the user (and return to Main_Page -> https://en.wikipedia.org/wiki/Barack_Obama -> https://en.wikipedia.org/wiki/President_of_the_United_States
First I'll change that so we test exact the same pages. But looking at the Obama page for now it's interesting. The first visual change is mostly the same or faster for the logged in user.
Hitting the Obama page without any browser cache
Hitting the Obama page with cache from the Main_Page
Hitting the Obama page as a logged in user
Summary
We can see that we get much more unstable metrics with items in the browser cache (we do 5 runs for all tests). I'm running these tests on DO so it would be better to also test on AWS. We can also see that there's not so difference in First Visual Change but a large difference in time spent in frontend vs backend (frontEnd loadEventStart - responseEnd and backEnd responseStart from the Navigation Timing API).
Next step
I would like to test the same URLs (without cache, hitting them each after each other and then also as a logged in users), run the tests for a week for three different locations outside of US and collect the metrics.
What about testing:
- São Paulo
- Mumbai
- Stockholm
And then we can collect the metrics and do a report.
