Hi,
Thanks for the detailed response. But wont it restrict the HTTP Origin to be a Amazon S3 bucket? In my scenario, I run my own Apache servers and I serve VoD directly from it if the client support. For other uses like RTMP or HLS, I use Wowza with MediaCache. So when a video is requested to Wowza and it hasn’t already cached, it will request the file via HTTP to my Apache servers, serve it to the viewer while the download is ocurring and when the download finishes, the whole file will be cache. So the next viewer that requests the same video will receive it directly from Wowza’s cache (the MediaCache).
The problem is Wowza is very limited and resource hungry. I have, for instance, no control over expiracy of objects. If one video is modified and I cannot change the filename, I have to restart Wowza to clean the cache, generating not only a downtime but also purging all other useful cached files.
Would be great if EMS could fetch VoD files from HTTP servers and cache them respecting the HTTP Expires header (that is used by the browsers and caching proxies like Squid to schedule the object expire time). That would allow a full featured CDN to be built with EMS+a HTTP server.