Scale-to-Zero Services

Idle services stop and cost nothing. The next request spawns them back.

How it works

Set idle_timeout in your config and tenement handles the rest. After five minutes with no requests, tenement kills the process and frees the memory. When a request arrives for that subdomain, tenement spawns a fresh instance, waits for the health check to pass, and proxies the request through.

[service.worker]
command = "python app.py"
health = "/health"
idle_timeout = 300              # stop after 5 minutes idle

Your app doesn’t need any hibernation logic. It just starts, serves requests, and exits when killed.

Measured cold wake times

These are real numbers from stopping an instance and immediately hitting its subdomain (measured on a MacBook, debug build):

Runtime	Cold wake (median)	Range
Python (stdlib http.server)	65ms	13-174ms
Node.js (http module)	105ms	104-110ms
Go (go run, cached compile)	140ms	100-224ms

Humans perceive anything under ~250ms as instant. These are all well under that threshold.

The economics

Most SaaS customers aren’t active at the same time. If you have 1000 tenants configured and 20 are active right now, the other 980 are costing you nothing.

Traditional: 1000 services always-on
  20MB per service = 20GB RAM
  Cost: 10 machines @ $500/month

Scale-to-zero: 1000 services, ~2% active
  20 running x 20MB = 400MB RAM
  Cost: 1 machine @ $5/month

The savings are roughly 100x, and the user experience is identical because the wake latency is imperceptible.