Today I setup my private instance using the EMI IE9/Chrome/Firefox/Safari - ami-8b2c2fff : 314854558937/ie9-20121031
Using a bash script I call the RESTful API 12 times every 5 minutes.
Some of the waterfall I have are like this ones, where is obvious the high CPU and the gaps.
So my doubt if call the RESTful API, on this EMI, in this way means that is too stressed and my waterfall are not real, so are affected by the high CPU on the test agent.
What size instances are you using? The small instances are pushing it (micro is out of the question) and Medium tend to be the sweet spot.
The latest test agents also add some additional logic to let the CPU calm down after launching the browser before starting the test.
I start with small and after restart with medium now it works like a charm…
I was trying to understand why on NewRelic I see people with high values on page rendering, while on the previous tests I did with the public instance of webpagetest I never have any issue on that. So I got always 2 sec on my tests but on NewRelic the average is 11 sec.
Thanks to the small instance I realized that some of the visitors to ours pages should be also “small”, in terms of the computers they use, so maybe they got bad values because of that, and for this reason the average on page speed was affected…
In any case now it works.
Thanks a lot.