A domain can configure its DNS settings so that multiple IP addresses are associated with it. When a client tries to resolve the domain name to an IP address, it returns a list of all IP addresses. Most clients would then send its requests to the first IP address in the list.

DNS load balancing is where the DNS changes the order of these addresses each time a new name resolution request is made. Most commonly, this is done in a round-robin manner.

Using this method, client requests should be distributed equally among all backend servers. However, load balancing at the DNS level has some major disadvantages:

• Lack of health-checks: The DNS does not monitor the health of the servers. Even if one of the servers in the list goes down, it will still return with the same list of IP addresses.
• Updating and propagating DNS records to all root servers, intermediate DNS servers (resolvers), and clients can take anything from minutes to hours. Furthermore, most DNS servers cache their DNS records. This means that requests may still be routed to failed servers long after the DNS records are updated.