A popular website can't run on one computer. Something has to decide which computer answers each visitor.
A busy restaurant on a Friday night. If every guest walked up to the same waiter, that waiter would drown while the others stood around. So a host stands at the door, glances at the room and sends each group to whoever has space.
A load balancer is that host for a website. It stands in front of several servers and shares each visitor out to one of them, so none gets swamped.
What it does
- shares the work: each visitor goes to a server with room to spare
- skips broken servers: if one stops answering, visitors go to the others
- makes growth easy: add another server and it starts taking visitors too
Why it matters
Without one, a site lives or dies on one computer. With one, a server can crash or be updated and visitors never notice. It’s one of the main reasons big sites stay up.
What it isn’t
It isn’t extra power on its own. If every server is already full, sharing visitors out more cleverly won’t help. You need more servers for that, which is scaling.
Where the comparison breaks down
A host might seat a few dozen guests an hour. A load balancer hands out thousands of visitors a second without the visitor ever seeing it. Some hosts also remember regulars and send them to the same waiter; load balancers can do that too.
How the words fit together
A load balancer sits in front of several servers. Adding more servers behind it is scaling. Skipping a broken server keeps uptime high. Sending visitors to a nearby, quiet server keeps latency low.
A load balancer also checks that each server is healthy. If one stops answering, it quietly stops sending visitors there until it's fixed.

When you'll hear it
“Put it behind a load balancer.” Run the site on several servers and let something share the visitors out between them.
“One of the servers went down but nobody noticed.” The load balancer sent everyone to the others, so the site stayed up.
“Traffic isn't spread evenly.” One server is getting more than its share and slowing down.
“If one of our servers fails tonight, does the site stay up?”


