# Cold Start

> A cold start is the brief wake-up delay the first request hits after an app has scaled to zero on Ownkube Compute.

**A cold start is the short delay a request experiences when it's the one that wakes an app back up after that app has [scaled to zero](/docs/glossary/scale-to-zero).** Because the app has no running replicas while idle, the first request in has to wait for a replica to start before it can be served.

When min replicas is set to 0 and an app has had no traffic for a while, Ownkube removes its running replicas entirely, so the app costs nothing while idle. The next request that arrives triggers a replica to start, and that request waits briefly while the replica comes up, then is served normally. Every request after that is served instantly, the same as always-on traffic, until the app scales back to zero from inactivity again. Keep min replicas at 1 or higher for anything that must answer instantly and can't tolerate the wake delay.

Cold start is the trade-off you accept in exchange for paying nothing while an app is idle. For low-traffic or personal apps, an occasional wake delay is reasonable in exchange for not billing at all between visits. For latency-sensitive production traffic, keeping at least one replica warm avoids the delay entirely, at the cost of some steady [working-set memory](/docs/glossary/working-set-memory) billing.

- [Autoscaling](/docs/features/autoscaling)

**Don't see a feature you need?** Email [support@ownkube.io](mailto:support@ownkube.io?subject=Feature%20request). Ownkube is shaped by the teams using it and we ship what our users ask for.