Calculator
What a fleet that follows your load would cost against the one you run today. An estimate; the Observe trial replaces it with your real numbers.
Nodes per hour, a peak daywith Hetscalepeak hoursaverage
00:0006:0012:0018:0023:00
every day: 6 h at the full fleet, 18 h at 3–4 nodesThe top of the chart is your fleet today; the bars are the estimate with Hetscale for one peak day. Every day looks like this.
1 · Money, per month, ex-VAT
Saving per monthup to€12
Pay for your normal load. Scale for your peak load.
- Compute today
- €51
- Compute with Hetscale
- from €34
- Hetscale plan
- €5
- Peak capacity
- 6 nodes, during peaks
- Average capacity
- 3.9–4.6 nodes
Autoscaling will not save money on a setup this small; Hetscale's value here is self-healing.
2 · Capacity and recovery
node-hours per month24–35 % fewer
today4,378 hwith Hetscale
2,828–3,344 h
utilisation today63–75 %
busy 3–4idle 2–3 outside the peak hours
a dead node at 03:00Replaced automatically
2–3 keep servingreplacement, health-checked before traffictoday: down until someone notices
New nodes deploy from a snapshot or through a webhook; Coolify has its own adapter.
3 · What this estimate cannot know
- Whether your app is CPU-bound. If it waits on I/O, CPU is the wrong signal and the estimate is optimistic.
- Whether the database is the bottleneck. More web nodes do not help a saturated database.
- The real shape of your peak. Three controls stand in for it here.
- Whether your fleet is bigger than your peak needs. The estimate takes today's fleet as the peak; if it is oversized, the real saving is larger. Observe measures it.
Replace the estimate with your data — connect read-only and get the report.
Connect read-onlyHow this is calculated
- Baseline nodes = ceil(fleet ÷ peak multiplier), never below 1.
- The traffic shape is three controls: peak hours per day, peak days (every day, weekdays 21.7 a month, weekends 8.7, or 1–10 a month) and how much busier the peak is, as a range (under 1.5× counts as 1.3×). The quick picks: Steady 1.3×, no peak hours; Daily peaks 1.5–2×, 6 h/day, every day; Spiky 2–3×, 2 h/day, every day; Campaign 2–3×, all day, 3 days a month. Hours above baseline per month = peak hours per day × peak days. Every peak day looks the same; the real hourly shape is Observe's job.
- During peak hours the group runs the full fleet you have today; outside them, the baseline.
- Each burst adds one started hour per node and peak day for the scale-out lead (threshold, provisioning, health check) and the scale-in window (threshold, cooldown, drain); Hetzner bills per started hour. That is 61–91 node-hours in this estimate.
- Burst node-hours are rounded up to whole hours because Hetzner bills per started hour; a month is 30.4 days.
- Baseline nodes cost the monthly price; burst node-hours cost the hourly price. Both are Hetzner's net list prices for the chosen location, 2026-09-17. Primary IPv4 addresses and traffic are not included.
- The Hetscale plan is the monthly Solo, Pro or Team price (ADR-0022); it is never per server. Extra groups, if any, are €4/month each.
- The low and high end of the peak multiplier give a range; the page shows the saving up to its high end and the autoscaled cost from its low end. This estimate: €34 to €40 compute, €6 to €12 saving.
- The load balancer is a prerequisite of every setup (ADR-0028) and, like volumes and snapshots, costs the same with or without autoscaling, so it is left out.
- Node-hours per month = nodes × 24 × 30.4 for the fixed fleet; baseline × 24 × 30.4 plus the burst node-hours with autoscaling. "Fewer" is the difference as a share of today's node-hours.
- Utilisation today = load ÷ the fixed fleet's capacity, averaged over the month: 1 ÷ peak multiplier outside the peak hours, 100 % during them. Idle nodes = fleet − baseline.
- Nothing here is a measurement of how fast a node is added or replaced; those numbers are published once measured.