The component that distributes incoming traffic across multiple backend replicas, the entry point for any horizontally-scaled service.
A load balancer (LB) sits in front of multiple replicas of a service and routes each incoming request to one of them. Algorithms include round-robin, least-connections, weighted, consistent hashing, and latency-aware. Cloud LBs (AWS ALB/NLB, GCP Cloud Load Balancing, Azure Load Balancer) handle TLS termination, health-checked routing, and DDoS protection out of the box. Service-mesh sidecars (Envoy, Linkerd) provide LB at the service-to-service layer.
Without a load balancer, every replica needs its own DNS entry and the client needs to handle failover, that's the 90s. With one, replicas come and go behind a stable address, health checks remove broken ones automatically, and the client doesn't know or care. Tuning the LB (timeouts, connection draining, slow-start, sticky-session policy) is a high-leverage SRE investment on every horizontally-scaled service.
See the part of the platform that handles load balancer in production.