Traditional DNS systems maintain a mapping of Internet Protocol (IP) addresses to fully qualified domain names (FQDN) in files called zone files, and those are usually text or binary format files that are maintained on one or more servers. Usually, the way the DNS operates is that we have one server that hosts the master DNS zone and one or more read-only replicas that host the slave DNS zone. The master is where we make the changes and the changes then propagate to the slaves through zone replication. This does provide high availability but if the master zone is disabled because of a failure, the DNS itself might become read-only so any updates to the DNS will need to wait for the master to recover.
One feature of the traditional DNS server types is that they are designed to update zones manually on the master server, and those that do have programmable access provide that access essentially on top of a traditional manually modified system. Another fairly common feature of the traditional DNS architecture is that the DNS servers themselves are not connected to the records they host; the servers will deliver a mapping between a name and an IP address, and all features beyond that operation are not included. For instance, if a request is made for a website that is currently not operational, the DNS will respond with the address but has no way of delivering any information on whether there is anything responding on the IP address provided.
In modern applications, however, we strive to deliver the highest availability possible within our application, and we also want to automate the provisioning tasks and deliver the website from multiple backends in case of hardware or other failures affecting our availability.
AWS identified that an application will be only as highly available as the sum of the components providing the features of the application, and one crucial feature is the DNS. Without the ability to resolve names to IP addresses, our application is essentially non-functional. And if the DNS service has an SLA of 99.9%, then we can theoretically never guarantee the delivery of any higher level of availability. To overcome the limitations of traditional DNS architectures, the engineers and architects at AWS decided to develop the Amazon Route 53 service.