I think it would be ideal if we could store our music on AWS and have the auto DJ play from it.  OR if we wanted to store the music at our home, our own network etc.

Basically have the option to store music somewhere else besides the machine that centovacast is installed on.

Can you explain the logic here?

Obviously the autoDJ is running on the server, and the autoDJ needs to be able to read the media in order to play it.  So you're saying... instead of uploading the media to the server... you want the autoDJ to read if from "somewhere else"?  Which essentially means downloading it from "somewhere else" when it needs to play it.  And instead of just keeping it on the server once it's downloaded it once, you want it to just play it, dispose of it, and then download it again every time the song needs to play?

I don't mean to belittle the idea -- it's possible that I'm misunderstanding or missing something here -- but it just seems like a huge waste of bandwidth to be re-downloading the same thing over and over.  It also seems a bit pointless to architect a "pull" solution when you could just "push" the media to the server as you can now.

Further, you run the risk of having your station go completely offline if your storage server becomes inaccessible due to networking issues/etc.

I guess maybe I just need a use-case to understand where this would be useful and the benefits would outweigh the drawbacks.

Ah OK, that makes sense.  I guess the big question is: which is cheaper these days, disk space or bandwidth?  Is it cheaper to pay the extra 5MB (give or take) per MP3 in storage space, or is it cheaper to pay to download that 5MB repeatedly over the course of the day/week/etc., every time it is played.  And I suppose that depends on how often the station plays each MP3.

If users were only using a subset of their MP3s it would be easy enough to just cache those MP3s and keep the rest on S3.  The problem is general rotation playlists -- if the user has 500gb of music in his general rotation playlists, we have to be able to play any of those tracks at any given time, which limits our ability to cache "most-used" MP3s (simply because there is no such thing as "most-used" -- in a random-playback situation, every MP3 will be equally utilized).

I like the concept, I'm just not sure if/how it could be reasonably implemented in a cost-effective manner.