Mirrored servers in data centers nationwide -- how do millions of users simultaneously get instant results from their "google.com" requests? I flunked my IT interview by getting this question wrong.
I thought that in the various metropolitan areas, an "http://google.com" request goes to the ISP's DNS server, which somehow returns an IP address for one of several geographically-nearby http servers, and then something internally rolls over to the next available local Google server. But then, I could not explain where the table of available local Google servers is actually cached, or the details of the IP address rollover. Or how they could manually take some server out of the rotation, from anywhere.
So, what should I be reading now so I can ace this question next time?
Also, what daemons run on these machines 24/7 to keep all those mirrored database disks synchronized?