Name or the Number! (2)

    This hopping around between server can take place any number of times but usually ends after two to three queries. Also, for this process to work every top level name server must know the IP addresses of all the root server beforehand. This means that this IP addresses must be coded into the server. Thankfully, the IP addresses do not change very often so this isn't a problem.
    The subdomain naming convention does not reflect the way computers are actually connected to the Net, since this is merely an identification system. Oracle for instance has a server named in.oracle.com that is physically located in India, not at Oracle Corporation's main site in the US.

Redundant Servers
  It is a good administrative practice to have more than one name server-one called the primary and the others redundant servers that download a copy of the primary server's database every once in a while, usually every six hours. These standby systems come into play when the 'primary' is either overloaded or goes off the network. There may be any number of secondaries as long as the administrator of the higher level domain (InterNIC at the root level) is aware of them.

Cache and load balancing
   One of the highlights of the domain name system design in 'caching'. The name server caches all the IP addresses for domain names that were requested recently, so that it can respond immediately if the same IP address is requested again. The cache is pruned frequently to remove old entries.
  Another upcomping use for name servers is load balancing. Large sites like www.msn.com have multiple IP addresses for the same domain name. The name server currently returns all the IP addresses, leaving the PC to choose one randomly. With new technology however, some name servers will now evaluate all the addresses to find the one with the least load, and then report it back to your PC.