I'm creating a REST API that will listen on a public URL and accept uploads of large amounts of data. I understand how to scale software processes using message queues, but what I don't understand is how to avoid a bottleneck between users and reaching my message queue server farm. To my understanding, at some point I have a machine listening to requests on an IP address, and that will be a single point of failure bottleneck. But this sounds really wrong since, you know, sites like google.com exist.
My subsequent question is if you can achieve this scalability for a fixed IP address instead of a fixed URL. Not because I have to do this, I just want to understand if in the practical sense of running a website (or other HTTP server) scalability can only be achieved by using a dynamic number of IPs.