Why load balancer?

Same fleet, same traffic. Pin every request to one server, then spread it across the fleet — including a mid-run outage.

Watch it explained

~24s animated walkthrough: traffic pinned to one server vs. spread across a load-balanced fleet.

Loading player…

Single fixed address

All traffic is pinned to Server 1. The other servers sit idle no matter how loaded Server 1 gets.

sent

success

—

errors

idle servers

Server 1 (pinned)

healthy

queue 0.0 · served 0

Server 2

healthy

queue 0.0 · served 0

Server 3

healthy

queue 0.0 · served 0

Server 4

healthy

queue 0.0 · served 0

Last routing decisionst = 0.0s

Run traffic to see routing decisions.

Why this matters

Capacity you paid for goes unused. Without a load balancer, everything funnels through one server while the rest of the fleet idles.
One dead server means an outage, not a dip. Kill Server 1 in "No load balancer" mode and every request fails — the client has no other address to try. Kill any server in "Load balancer" mode and traffic just reroutes.
The algorithm matters under skew. Round robin assumes uniform requests; least-connections adapts when some requests are heavier than others.

Traffic console

Run traffic, then take a server down mid-flight.

total traffic (req/s)Total client traffic in requests per simulated second, regardless of how many servers are up.server capacity (req/s)How many requests per second a single server can drain. Past 2× this, a server is treated as overloaded and starts dropping requests.fleet sizeNumber of servers in the fleet. Resets the simulation.simulation speedOnly changes animation/simulation playback speed.

Try this

Run traffic until Server 1 turns red, then click its power icon. Every new request fails immediately — there's no fallback address.