Experienced advices for high traffic website / CDN's with origin pull feature ?

Hello guys,

Recently i got into rackspace and built a stack based on nginx / unicorn / rails / mongodb / ubuntu and so far am really happy with it.

Now am starting to change my focus from development to testing and improving performance, so i started by trying to give a change to Cloud Files but perhaps it still have one annoying inconvenient ( at least for me, let me know if there’s a nice workaround for those ) which is lack of the origin pull feature.

Perhaps going for Amazon Cloudfront or AKAMAI will solve that issue smoothly ( am i wrong? any advices on this ? ).

Another thing is, my interface is built on top of another framework wich isn’t integrated into sprockets, its MVC, has it own compiler, and finishs delivering me one nice and neat css and js file.

My idea so far is to refresh the cache for those assets by creating a folder with an auto increment number for each deployed version, i mean, i wish i had a better integration with sprockets here but so far this is the easiest way i could think about.

The site will have huge amount of visits, and perhaps the only “dynamic” feature i’ll have is search, so my idea is to keep only one server with a reasonable amount of RAM here in london, and cache everything else on the CDN.

As this is my first application written in rails i would like to know if any of you guys have anything to add or had experience problems before with this sort of strategy.

Thanks a lot.

respect,

peace