Every now and then I like to look through the test history to see what kind of opportunities for improvement there are. This morning I saw what looked like the most painful example of why persistent connections should be used: [url=http://www.webpagetest.org/result/090417_1CQ5/]http://www.webpagetest.org/result/090417_1CQ5/[/url]
74 https requests all without persistent connections or expires headers (and in this case hosted in Japan and tested from the US). Now granted, I doubt you’ll be ordering pizza from a Pizza Hut if you’re not somewhat close to their location but even just negotiating all of those extra SSL connections has to hurt.
As it is, it took 34 seconds to load over FIOS for the first view and only came down to 24 seconds for the repeat view (which was only 37KB of data!!!).
Hopefully it’s just a testing harness since it’s not on the pizzahut domain and they get it fixed before the deployment but OUCH!
We have been checking our site using your tool and have found it very useful! We do have a question regarding the results related to Keep Alive. We are running Apache with a Proxy to Lighttpd for static files. Both Apache and Lighttpd have Keep Alive turned on we see Keep-Alive in the response headers. Below are the Response Headers for two different files. One being served by apache and one being served by lighttpd. The tool consistently shows that Keep Alive is not working for any of the files we are serving. I was hoping you could shed some light on what we might be doing wrong or any changes we can make to our webserver configuration files to make sure that Keep Alive is being utilized effectively.
Here is a link to a recent test we have done. Thanks in advance for any information you can provide!
Do you have a proxy or anything sitting in front of it that could be closing the connection or a browser-specific rule controlling it (turning off persistent connections for IE8)?
Thanks for the response. After doing some checking the Apache conf file had the following line
SetEnvIf User-Agent “.MSIE.” nokeepalive ssl-unclean-shutdown
My guess it was there due to older version of IE related to SSL certificates.
Once I removed that line and re-ran the test the results show that it is working.
Thanks so much for the suggestion and creating such a useful tool!