Load Balancer & Request Queueing Simulator

Traffic

Arrival rate λ 12 req/s

Backend servers 4

Capacity skew (weighted) 30%

Routing Algorithm

Failover

Live Metrics

Throughput

0.0 req/s

Avg latency

0 ms

Max queue depth

Dropped requests

How it works

Each backend server is an independent queue: requests wait in line (FIFO) and are served one at a time at rate μᵢ (its service capacity, in requests/second). Arrivals follow a Poisson process with rate λ — inter-arrival gaps are drawn as -ln(rand)/λ, the standard way to sample exponential inter-arrival times.

Little's Law:  L = λ · W
  L = avg. requests in the system
  W = avg. time a request spends in it (queue + service)

Utilisation:  ρᵢ = λᵢ / μᵢ   (per server; ρᵢ ≥ 1 ⇒ queue grows unbounded)

The load balancer decides which server gets each incoming request:

Round Robin — cycles through healthy servers in fixed order, ignoring current load.
Least Connections — sends each request to the server with the fewest requests in flight (queued + in service); adapts automatically to heterogeneous or failed servers.
Weighted Random — picks a server randomly with probability proportional to its capacity μᵢ, so faster servers statistically receive proportionally more traffic.

Every server queue has a finite capacity; a request that arrives when its target queue is already full is dropped (shown in red) — exactly what happens when a real service returns 503/overload under a traffic spike. Toggling Fail Server 1 removes it from rotation entirely, so you can watch each algorithm's failover behaviour: Round Robin keeps trying to skip it, Least Connections and Weighted Random reroute its share onto the survivors.

Request in flight Completed Dropped (queue full)

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.