CDN experiment: Free CDN shockingly faster than Amazon CloudFront in many cases
speedtestdemon.com
speedtestdemon.com
At some point we considered multiple downstream providers and balancing between them in a geo distributed way. If a provider was slow or down in a geo, it'd automatically switch to PoPs from a different provider who would be faster at that time.
Even this doesn't really do it though, if you like to manage your config and infra as code, you'll have to support different configurations across different providers, and having to always review and deploy multiple sets of changes that need to maintain parity. CDNs also have different behaviors in how they deal with stampedes and things like that, plus you need to manage multiple vendor relationships.
I've come to the conclusion that I'd rather hitch my horse to the Cloudfront cart, ride their availability while paying up for the privilege and at some point, if the economics make sense and we have people with idle hands, start running our own CDN.
I'm pretty sure Cloudfront only compresses text, so no idea why they are comparing compressed images.
Riddled with factual errors. This is not a good article.
I'd like to see an analysis of more CDNs before trusting something like this.