If the Internet were small or centrally controlled, that could be done using a database on a central server, but that does not work with a huge, rapidly growing network. (In computer jargon, that solution does not scale well).
Instead, we use a decentralized system in which the administrator for each domain maintains a local database of names and corresponding IP addresses on a domain name server (DNS). (This is our second meaning for "DNS" -- it also stands for domain name system).
Every domain must have at least one DNS. For example, on our campus, we have a DNS that keeps database of host names and their IP addresses for the csudh.edu domain.
Like the domain name system, the domain name servers are hierarchical. When a program wants to find the IP address of a host, it calls its local DNS server. If that server does not have the information, it passes the request up to a higher level server.
For example, a DNS request for the name sws.csudh.edu could be resolved by the DNS on the CSUDH campus, but the name sws.mit.edu would be passed to a DNS for the edu domain. That DNS would pass the request on to the DNS at MIT. The MIT DNS would look up the IP address, and pass it back to the host that had requested it.
If it were not in the MIT DNS, an error message would be passed back. The IP address would also be cached in the CSUDH DNS for speedy retrieval if it were needed again.
A request to a different top-level domain would be passed through the root servers that link them together. For example, the root server know the location of the servers for the edu and com domains.
