Before a client can connect to api.example.com, it first has to ask DNS for an IP address. DNS load balancing uses that lookup as a traffic-routing decision.

Instead of always returning the same address, the authoritative DNS service can return different answers for the same hostname. It might choose an address based on region, health, weight, latency, or failover state.

This makes DNS useful for big decisions: send users to a nearby region, shift a small percentage of traffic during a migration, or move new lookups away from a failed region.

The tradeoff is that DNS is not precise. DNS answers are cached by resolvers and clients, and DNS cannot see what is happening inside your servers. It does not know which backend is least busy, which cache is warm, or which GPU worker has room for one more request.

This chapter explains where DNS load balancing works well, where caching limits it, and how production systems usually combine it with real load balancers behind the DNS answer.