When a site is too busy for one machine, or needs to stay up if one machine fails, several identical servers run the application and a load balancer sits in front of them. It receives each request and forwards it to a healthy server, using a method such as round robin or least connections. It checks the servers regularly and stops sending traffic to one that is not responding.
Load balancers can work at different layers. A layer 4 balancer forwards connections without reading them, while a layer 7 balancer understands HTTP and can route by hostname or path, terminate TLS and add headers. Cloud providers sell them as managed services.
Because every request passes through it, a load balancer is also a natural place for security controls such as TLS, rate limits and a WAF. It is also a limit: it may enforce its own maximum for request size, header size or timeouts, and these limits shape what the servers behind it ever see.
Know what sits between the internet and your servers, because those components decide which requests arrive and what the logs show.