sorry if this should turn out to be a silly question, but I am wondering about setting the numbers of tests to more than 1, and how that influences tests results.
If I check my website in Network Console for Chrome or Firefox, it loads snappy, usually under 2.5 seconds, with a TTFB of 750-1000ms.
All other major tests tools seem to confirm that performance. As does running a single WebPageTest test.
If I set the number of tests to run to 3,4,5 or more, suddenly the results change. The first test result usually is fast, but the other ones tend to get increasingly slower, with TTFB’s of up to 3 seconds, as seen here: http://www.webpagetest.org/result/140812_4S_11WG/
This does not always happen though, sometimes it stays fast and snappy: http://www.webpagetest.org/result/140812_QJ_T4K/
So I am wondering, does running all these tests simultaneously affect the test results through unusual server stress levels? Should I dismiss this and simply stick to running a single test, or should I worry about my VPS not being able to handle 5 simple performance tests?
Please also note that this mainly happens on the Dulles test server, I just tried Los Angeles, no issues.
My host could not find any anomalies on my server, or slow SQL / PHP issues, yet these test results make me worry a lot - too much maybe?
Any feedback or thoughts would be welcome and much appreciated.
Thanks a lot in advance,
Dulles has several test machines and the tests are run simultaneously if enough are available so it is most certainly generating more concurrent traffic than a single test at a time (most other locations are only a single machine). That said, the limit of 9 runs is tiny as far as load testing goes so I’d have a strong, hard look at your backend (and yes, you are very right to be concerned).
I’d strongly recommend installing New Relic or another APM solution which will instrument your back-end and tell you where the time is going. Considering you are seeing long times even serving static resources I’d look at the web server config first. An easy fix might be to put varnish or Nginx in front of Apache to serve the static files and only have Apache handle the heavier php requests (or switch to Nginx outright - though I have a bad history with scaling Apache and may be biased).
thanks for your feedback. It confirms what I was suspecting. Most single tests like Pingdom FPT, Gtmetrix or even using the Network Console for Chrome or Firefox return rather fast results, but once in a while the page hangs, probably when there happen to be multiple simultaneous requests. I don’t have more than 40-60 hits per hour, so probably most of the times there is not much stress on the server, hence a good performance most of the time, and then boom, suddenly it really lags.
I will contact the professional services team of my host to check the server settings, any recommendations what I should ask them to do? I use Prestashop, and other Prestashop users have recommended checking if the LAMP stack is optimized, and PHP / SQL caching.
And like you said, what really makes me wonder is why images are often served so slowly, from a TTFB point of view, there’s no reason. Any recommendation on what to look for to fix that?
Thanks a lot again, your help is much appreciated!
All the best,