Work out requests per second capacity instantly with clear inputs, formula shown and shareable results.
Throughput capacity is threads divided by service time — 64 threads at 45 ms each gives about 1,422 requests per second per instance at full saturation. Running at a utilisation target below 100 percent keeps latency stable, and checking capacity with one instance removed is what turns a capacity number into an availability guarantee.
Throughput capacity
per instance max = threads x 1000 / service time in ms; safe = max x utilisation target; fleet = safe x instances
Queueing. As utilisation approaches one, waiting time grows as rho/(1-rho), so the last 20 percent of nominal capacity costs disproportionate latency.
N-1 at minimum, and N-2 across two availability zones. Sizing for N means any single failure causes an overload.