| A critical component of securing an information system is ensuring its availability. The best way to ensure availability is to eliminate any single point of failure in the system itself and in the network architecture that supports it.
Fortunately, the inherent design of DNS supports a high-availability environment. Master and slave servers regularly communicate zone information, so if any name server is disabled at any time, another can immediately provide the same service. The task for the network architect is to ensure that a disaster or outage cannot simultaneously impact both the master and all of its slave servers. If a disaster occurs, the DNS protocols cannot prevent total loss of name resolution services for hosts within affected zones.
The solution is to disperse name servers in such a way as to avoid single points of failure. At minimum, authoritative name servers for the same zone should be on different network segments in order that at least one name server is available in the event that a router or switch fails. This fault tolerance should also extend to wide area data communications lines. For example, if a site has multiple leased lines connecting the network on which the name server resides to a larger network such as the NIPRNet, routing protocols should be configured such that if one of the lines fails, another one will still be available to support the name server.