Load Balancer

Step by step

The example above, written out — the same steps the animation plays.

1. 1Round robin. Round robin sends requests to A, B, C, A, B, C… in turn, whatever they are doing.
2. 2#1 (slow) → A. A: next in turn. Every slow request lands on A, because they arrive every third tick — exactly in step with round robin.
3. 3#2 (slow) → B. B: next in turn.
4. 4#3 (slow) → C. C: next in turn.
5. 5#4 (slow) → A. A: next in turn. Every slow request lands on A, because they arrive every third tick — exactly in step with round robin.
6. 6#5 (slow) → B. B: next in turn.
7. 7#6 (slow) → C. C: next in turn.
8. 8#7 (slow) → A. A: next in turn. Every slow request lands on A, because they arrive every third tick — exactly in step with round robin.
9. 9#8 (slow) → B. B: next in turn.
10. 10#9 (slow) → C. C: next in turn.
11. 11#10 (slow) → A. A: next in turn. Every slow request lands on A, because they arrive every third tick — exactly in step with round robin.
12. 12#11 (slow) → B. B: next in turn.
13. 13#12 (slow) → C. C: next in turn.
14. 14Done. Most work queued on one server: 9 ticks. Round robin balances the number of requests, not the work.

What's happening?

1. Round robin hands requests to each server in turn — fair in count, blind to how heavy each request is.
2. Least connections sends each request to the server with the fewest requests in progress, so slow work spreads out.
3. When a server stops answering, health checks mark it unhealthy and traffic goes to the others.

Where you'll meet it

Every site with more than one server: AWS ALB, NGINX, HAProxy, Kubernetes Services. It is also what makes rolling deploys possible without downtime.

FAQ