Well, he said large-scale, but 500 users a day is very small. If he had 500 concurrent users, I would say something like that could need a mid-sized server, depending on the infrastructure needs. There's really not enough info here in the thread to make a good determination.
What I can say, though, is that AWS is probably unnecessary except in extreme circumstances. It's an unnecessarily expensive solution.
For a good recommendation, we'd need to know what kind of files, whether they can be compressed into a handful of files/1 file. How large of files.
An example, could you replace them with HTTP requests that return an xml file, compressed with gzip?
I could get you a server that could do that using nginx, serving 15k simultaneous requests at a time for maybe 100 bucks/mo, depending on a lot of factors. Need redundency? Double or triple that price. Need more? Get your own server and a dedicated line.
Need 1 15mb zipped file to be sent to 10 simultaneous users only when the user opens the program? 10/mo tops, with redundency? 20 bucks. Want it stored in-house? 100 dollar second-hand server and a 100 dollar contract with your cable company, and you can do other stuff with the left-over bandwidth.
