Each server rack's height is driven live by its queue depth, and its instantaneous
service rate degrades under contention (queueing-theory style, like M/M/1):
rate(q) = capacity / (1 + k·q)
wait time ≈ q / rate(q)
- Round Robin — cycles through racks in fixed order, ignoring current load or capacity.
- Least Connections — routes each packet to whichever rack currently has the shortest queue.
- Weighted (smooth WRR) — capacity-aware, like nginx's weighted round-robin.
- Arrival rate — packets/sec launched from the mobile-client swarm above the scene.
- Server instances — how many rack columns share the load.
- Heterogeneous capacity — unequal rack sizes; shows how Round Robin ignores capacity while Least-Connections and Weighted adapt.
Real use: cloud load balancers (AWS ELB, GCP LB, Kubernetes Services) route mobile-app API traffic across backend replicas with exactly these algorithms.