Some tests in my private instance product TTFB more than Start Render. How is that possible? Are these reliable test results? What should I look for?
What kind of an app are you testing? Is there some kind of offline behavior built into the app?
It is a simple website. www.walmart.com. Few times I have seen on other urls as well
eg. Sensodyne Full Protection Whitening Sensitive Toothpaste, 4 Oz - Walmart.com
I cannot reproduce it easily since it happens once in a while and only in our private instance. Cannot paste the test link (since its private instance) but waterfall is pasted as a screenshot.[attachment=833]
was your test running on the same machine, same browser all the time?
Remember JS, CSS execution time depends on client machine’s CPU and memory.
I also notice that walmart’s base page is running on HTTP/2 protocol.
Does walmart still have HTTP/1.1? If yes, you might want to try if this can be reproduced on HTTP/1.1.
But my main concern is that why the page stays blank for 2 seconds based on my test below.
there are 20 assets contribute to the blank page.
there is a huge JS file (almost 0.5MB despite it’s already compressed) on line 10 taking 1.5s to download + execution time.
I would suggest you profile all JS files that are in the the first 2 seconds.
I suggest to use Windows Performance Toolkit (free from Microsoft) to do JS profiling and do it on a medium size machine with warm cache.
Interesting situation. When you ran the test did you have video capture enabled? If so that might shed some light into what’s happening at 0.7s vs 1.6s, ie what is actually ‘rendering’ at 0.7s?