A popular service handles more requests than any single server can manage, so the work is shared across many. A load balancer sits in front and distributes incoming requests among them, aiming to keep each server busy but none overwhelmed.
It also notices when a server becomes unhealthy and stops sending it work, so a single failure does not reach users. How it decides where to send each request, by turns, by current load, or by which server already holds related data, shapes both performance and reliability.