Watch requests distribute across servers in real time
Requests are distributed sequentially across servers. Simple and effective when servers have equal capacity and request cost is uniform. Doesn't account for server load or request duration.
The load balancer sends periodic HTTP requests (or TCP pings) to each server. If a server fails N consecutive checks, it's removed from rotation. Traffic is automatically redistributed to healthy servers.
Without sticky sessions, each request may go to a different server — meaning session data (shopping cart, login state) stored in-memory would be lost. Solutions: sticky sessions (same server), or shared session store (Redis).
GeoDNS returns different IP addresses based on the user's location. Anycast uses BGP routing — multiple servers share the same IP, and internet routing protocols direct users to the nearest one automatically.