11–15 of 15 posts
I can't tell how similar this is to CloudFlare, but it looks close. That said, CloudFlare actually increased latency by a non-trivial amount, which caused me to abandon it. It may be better now, or it may have worked better on much larger sites, but my 256M slice & Django blog was faster without it.
I can't tell how similar this is to CloudFlare, but it looks close. That said, CloudFlare actually increased latency by a non-trivial amount, which caused me to abandon it. It may be better now, or it may have worked better on much larger sites, but my 256M slice & Django blog was faster without it.
CloudFlare as I understand are focused on being a really good and easy to use CDN.
And yes, the cloud approach means that we add about 200-300ms to every non-cacheable request. We think it's a decent trade-off since we usually shave off 10 times that when all our optimizations are applied.
I think Google has a similar tool freely available, although I can't find it right now. Does anybody know?
One reason people may prefer a cloud solution is there is no other infrastructure to set up. It turns out mod_pagespeed is both CPU and cache intensive which means rolling this out on your servers requires beefing up your server infrastructure.
We may end up using their technology & contributing to the project at some point, but right now it's very alpha and breaks a lot of sites (including Google's own!).
They're also missing most of the browser-specific optimizations that we're doing, specifically around mobile.
Hi bmelton, Looks like you were active around the TC Disrupt time, which was very busy for us (we've added two datacenters since then). We generally make websites about 30% faster on average.
The gig was to migrate some of his company's products to the web, and of course, his main concern was on the speed -- he didn't believe the web could be as fast as a native client. While there's merit to that on its own, my attempts to show otherwise were thwarted by a really slow CloudFlare CDN, at the time, so I kiboshed it.
I blogged about it at the time, and hoped that it was just due to what I'm sure was unexpected expansion in the light of the Disrupt coverage, but truthfully, I don't currently have any products large enough to benefit from it.
It seemed like a cool service, and having recently gone the other way (implemented Amazon's CloudFront / caching / yadda yadda), I will attest that CloudFlare's implementation was far easier and more convenient. I'll try to check you guys out again soon.
Thanks for the comment!