I’m working on a friend’s Wordpress site that uses W3TC and MaxCDN, trying to get it to run faster. I’ve been getting insanely slow results. In fact, disabling all the caching in W3TC makes the site faster, not slower.
WebPageTest does not detect the CDN weather it’s on or not
I’d be the first to admit that I don’t really know what I’m doing, but I’ve followed several sets of instructions of getting W3TC and MaxCDN working together, and I don’t understand why I’m seeing no improvement (and in fact am seeing the opposite).
Wow, those first byte times are insanely high. The page cache can’t possibly be working when W3TC is turned on (I know a few people have seen similar issues). I believe it requires changing some directory permissions when you first install the plugin - do you know if you did that step?
You can enable the W3TC debugging which will dump a bunch of information to the page source (as a comment at the end of the page when you view source). The debug information can also be useful because it will show you the database queries and how long each one takes which is helpful for identifying a specific plugin that is making things slow.
If you’d like me to take a look and poke around feel free to ping me off-board (pmeenan@webpagetest.org) and I can take a quick look to see if it’s something easy.
As far as the CDN goes, when it is enabled there are a lot of requests that aren’t being served by the CDN (possibly coded directly into the template?) but there are also a fair number that are. To see the benefit from the CDN you need to do several runs to let the CDN cache the content first. That said, it looks like your web server is relatively close to the Dulles test location so the improvement would actually be the smallest there - you should see larger gains from California or Europe (though not worth looking at until the first byte times are fixed).
Not sure if rellated but I am also using W3TC / MaxCDN and Cloudflare as well. I am loading my homepage in under 4 sec using the Dulles test site but even changing different settings I am getting 48-51 sec load times in SF! ???
WebPageTest Test - Running web page performance and optimization tests... Results are better from the same location using Chrome but IE 9 just kills my page in SF. What is even stranger is Page Speed gives me a Page Speed 1.12 Score: 93/100 even though the page takes almost a minute to load.
Page runs like a champ out of my default location (?) Dulles. I am physically in north NJ.
Holy cow, I ran a couple of tests just in case it was a problem with the agent but it looks consistent and testing other sites from the same agent looks fine. It’s bizarre that it’s only the base page socket connect that is suffering but it’s taking an insane amount of time to establish the connection. It looks a lot like packet loss but I’d expect that to show up in other connections as well and not as consistently.
Do you have any IP-based rate limiting, security modules or anything else that may be holding down the connections? Does the server have enough clients/file descriptors?
Your actual page performance isn’t nearly as bad as the measurement - if we can figure out why the socket connect is taking so long the numbers should be a lot closer.
Do you have any IP-based rate limiting, security modules or anything else that may be holding down the connections? Does the server have enough clients/file descriptors?
<<< I couldn’t tell you that myself - I am just on a shared-hosting plan and am not IT by profession. I’ll ask my host about this.
I asked my web host about this and he reminded me they just upgraded to a faster new server at LiquidWeb and that they arent even scratching the surface for using resources - he suggested that maybe Cloudflare has something to do with it so I will check that out - thx
Ahh, didn’t realize you had Cloud Flare in place. They are working on whitelisting the test agents but you can bypass their filtering by turning down the security controls in their control panel. That would certainly explain it.
Time to first byte is insane at SF. I suspect that this is not on my end, since all the other tests world wide are fine. I’ve tested the exact same page from all available locations. I can provide the other test URLs for comparison if needed.
Thanks, I’m working with Strangeloop to update the old San Jose IE7 location to be IE9 so I’ll probably ping fastsoft and take the SF agent offline (next day or two).
The San Jose agent is now online with IE 8, 9, Chrome, Firefox 3.6 and Firefox 7. Sorry about the issues with the San Francisco agent (which I also took offline).