Autoscaling

Autoscaling automatically adds or removes compute capacity in response to real-time demand, so your app keeps up during traffic spikes and stops paying for idle machines when things go quiet. Horizontal autoscaling spins up more identical instances behind a load balancer; vertical autoscaling resizes an instance to a bigger machine. Rules trigger on metrics like CPU, memory, request rate, or queue depth. Why it matters: for a SaaS with uneven load — a Monday-morning rush, a launch on Product Hunt — autoscaling is the difference between a graceful ramp and a crashed site or a wastefully over-provisioned fleet. Practical note: scaling is not instant. New instances take seconds to minutes to boot and warm caches (see cold start), so set thresholds to scale up early, not at 100% CPU. Set a sane maximum to cap runaway bills from a bug or bot flood, and load-test your scale-up path before you actually need it.