At my new job I’m working on a web app, which isn’t too far off from the typical deployment of modern web applications: someplace somewhere (who knows?) there’s a server farm, but that server farm isn’t running our application. Instead, they’re running a distributed (so that any one processor can go down and nothing is lost) instance of an emulator, which is running a whole bunch of virtual servers. These virtual servers are also distributed, so that any node can go down and nothing is lost. On this virtual distribution of servers the applications reside – but they’re not just running the services directly! Nope! Instead, the services are distributed among a host of applications, so that individual services can be troubleshot or upgraded individually, but even that’s not virtual and isolated enough. Inside any given application, the services aren’t run directly, but instead are run in separate, protected threads that are all surrounded by error handling protocols, and there the service is finally allowed to run. And this is all standard practice!
It’s layers upon layers of abstraction, to the point where any punch card engineer would blow a gasket, and probably just break down and cry. I think it’s awesome.