One address, several service locations
Multiple nodes advertise reachability for a service address, allowing users to access it without first obtaining every node’s individual address. Different regions can consequently reach different serving nodes.
This differs from multicast replication and from DNS rotation among different addresses. The techniques can be combined, but their selection layers and semantics remain distinct.
Proximity in routing terms
Routing and policy form the regions from which each node attracts traffic. Interconnections, attributes and failures influence selection, so nearest cannot simply mean shortest straight-line distance.
Adding a node does not guarantee equal load. Client-side routing decisions and topology can concentrate traffic, requiring capacity to be understood together with actual catchment areas.
Route changes do not copy session state
A route change can send later packets to another node. Stateful services therefore need additional arrangements: sharing an address does not synchronize connections or application data.
Application failure also needs coordination with reachability. A node can keep advertising an address while failing to serve requests; health detection, route withdrawal and data consistency remain operational responsibilities.