One service, many backends
A front door can distribute the work.
A load balancer presents a service endpoint and forwards requests to backend servers. The client usually does not need to know which backend answered.
Distribution can use several strategies, such as round robin, least connections, weights, or session-aware rules. A health check helps prevent new requests from being sent to a backend that is not ready.
Build the model
Four balancing pieces
Front door
The endpoint that receives requests for the service.
Backend pool
The servers available to handle requests.
Health check
A test that identifies whether a backend can receive work.
Distribution rule
The policy used to choose among healthy backends.
Mini experiment
Choose a backend
Select the pool state and see which backend should receive the next request.
The next request is sent to Backend A because the round-robin turn is ready and both servers are healthy.
A healthy pool can share requests across both backends.
Knowledge check
Test your load balancing model
Answer five questions. The answer order changes each time.