3 ms·

I'm not surprised he got these results. While the author raises some valid points there are a lot of flaws with this. He doesn't say what the times are for. Is

I'm not surprised he got these results. While the author raises some valid points there are a lot of flaws with this. He doesn't say what the times are for. Is it the time to DOM ready? To the page completely loaded and rendered? Did he load each site at least once before running the tests to make sure CF's caches were primed? CloudFlare can make your site faster or slower depending on how you use it and what you're serving through it. Since it's a proxy response times will be higher for dynamic content it needs to fetch (the page itself, unless it's cacheable). Static content (with proper cache headers) will be served through its CDN (everything else) which will almost always be faster, again assuming CF's cache is primed. The author's finding may be accurate about a base level install of these platforms with no performance improvements. I believe that without proper caching headers CF is going to have to query your server for everything to make sure nothing has changed. Sending proper cache headers with your static files will eliminate this issue and improve your performance. AFAIK none of these platforms use proper caching headers out of the box. One last nitpick. Why the hell is this article spread across 6 pages? That's incredibly useless, annoying, and takes away from the user experience. Edit: Originally I called out the author for claiming his site has a PageSpeed score of 97. I did so because the site issues 57 requests, including 8 CSS and 20 JS files, some of which aren't even minified. I assumed there was no way his score could be 97 with these glaring issues. Turns out I was wrong. I tested it[1] and it does have a score of 97, which just goes to show that PageSpeed and YSlow have plenty of their own issues. IMO his site shouldn't have over a 90 based on these obvious and easy to fix flaws. 1: http://www.webpagetest.org/result/130105_G4_DEA/ http://www.webpagetest.org/result/130105_G4_DEA/

xpose2000 14y ago
I don't really understand Cloudflare's problems with latency. When using FULL CDN+Optimizations the entire site's HTML and assets should be cached and served super fast via cloudflare. That is the best case scenario for them to speed up load times. But based on my testing, Cloudflare adds more latency to response times even in a best case scenario of full caching. However, when I do the same exact type of optimizations with Google PageSpeed service I see great results. The most interesting part about cloudflare's problems is that they weren't there 2 years ago when I first started using it. I blogged about how great it was. Here are the articles I've written about cloudflare in the past 2 years: http://www.x-pose.org/category/cloudflare/ http://www.x-pose.org/category/cloudflare/
saurik 14y ago
> Since it's a proxy response times will be higher for dynamic content it needs to fetch (the page itself, unless it's cacheable). This is actually not true; I mean, it might very well be true for CloudFlare (as they may be doing something exceedingly stupid at the edge, might have bad connectivity for their servers far from the trunk, etc.), but it does not follow in general: due to how TCP works, there are numerous advantages to both throughput and latency from adding an intelligent middle-man. For reference: http://news.ycombinator.com/item?id=2823268 http://news.ycombinator.com/item?id=2823268 http://news.ycombinator.com/item?id=4203371 http://news.ycombinator.com/item?id=4203371
- driverdan 14y ago
  Interesting point. This will be dependent on the site receiving significant traffic will it not? If the CDN node you go through doesn't have an open connection it will be slower. If that CDN node isn't getting new traffic it's going to close it. It's only going to hold that connection for a reasonable length of time.

xpose2000 14y ago

I don't really understand Cloudflare's problems with latency. When using FULL CDN+Optimizations the entire site's HTML and assets should be cached and served super fast via cloudflare. That is the best case scenario for them to speed up load times. But based on my testing, Cloudflare adds more latency to response times even in a best case scenario of full caching. However, when I do the same exact type of optimizations with Google PageSpeed service I see great results. The most interesting part about cloudflare's problems is that they weren't there 2 years ago when I first started using it. I blogged about how great it was. Here are the articles I've written about cloudflare in the past 2 years: http://www.x-pose.org/category/cloudflare/ http://www.x-pose.org/category/cloudflare/

saurik 14y ago

> Since it's a proxy response times will be higher for dynamic content it needs to fetch (the page itself, unless it's cacheable). This is actually not true; I mean, it might very well be true for CloudFlare (as they may be doing something exceedingly stupid at the edge, might have bad connectivity for their servers far from the trunk, etc.), but it does not follow in general: due to how TCP works, there are numerous advantages to both throughput and latency from adding an intelligent middle-man. For reference: http://news.ycombinator.com/item?id=2823268 http://news.ycombinator.com/item?id=2823268 http://news.ycombinator.com/item?id=4203371 http://news.ycombinator.com/item?id=4203371

driverdan 14y ago
Interesting point. This will be dependent on the site receiving significant traffic will it not? If the CDN node you go through doesn't have an open connection it will be slower. If that CDN node isn't getting new traffic it's going to close it. It's only going to hold that connection for a reasonable length of time.