Load balancer
A load balancer distributes incoming traffic across multiple backend servers using different algorithms (round-robin, least connections, IP hash), to maximize availability and prevent any single machine from becoming a bottleneck under high load.
In practice
Nginx and HAProxy are the most widely used open-source load balancers for self-hosting; public clouds offer managed options (AWS ALB, GCP Load Balancing). Health checks are the critical feature: the load balancer must detect when a server stops responding and automatically remove failing instances from the pool. Without health checks, a failing instance continues receiving 25% of traffic and degrades one in four user sessions.