Load balancers sit in front of a pool of backend servers and forward each incoming request to one of them based on an algorithm: round-robin (sequentially cycles through backends), least-connections (sends to the backend with the fewest active connections), IP-hash (always routes the same client IP to the same backend for session stickiness), or weighted round-robin (favours higher-capacity backends).
For scraping platforms, load balancers distribute API requests across multiple API server instances, distribute scraping jobs across worker pools, and route browser sessions to available browser farms. Health checks ensure that failed backends are removed from rotation automatically.
From a scraping target's perspective, a load balancer means that consecutive requests from the same IP may reach different backend servers, which can affect session-based anti-bot systems that store state on individual servers. Sticky-session load balancers address this by ensuring a given IP or cookie always routes to the same backend.