Two quotes arrive: self-managed VPS at €12/month, or "managed server" at €89/month with monitoring and patches. Your CTO calculates quickly: €77 × 12 ≈ one freelance ops day. Except managed does not include application migration, PHP tuning, or debugging your Friday evening deploy — and self-managed without on-call sometimes means twelve hours downtime before intervention.
Managed vs self-managed plays out in real euros per hour, not "peace of mind" slogans.
What managed removes — and leaves
| Task | Often managed | Stays with you |
|---|---|---|
| OS / hypervisor patches | Yes | Applications |
| Infrastructure snapshots | Sometimes | Application restore |
| Host network firewall | Sometimes | App firewall, auth |
| 24/7 on-call | Rare at entry level | Business incidents |
| Compliance | Partial documentation | DPA, architecture |
Paying managed without reading scope is buying a support FAQ — not outage insurance.
Hidden cost of self-managed
CVE patch time and planned reboot. Monitoring setup (uptime, Prometheus, logs). Implicit on-call from senior developer. Config errors: untested backup, overly open firewall.
If nobody is designated "ops", self-managed is a variable-rate loan in incidents.
Quick decision formula
Estimate ops hours/month × loaded hourly cost. If hours × rate > managed margin and provider documents restore + on-call → managed rational. If existing ops team + root/custom kernel need → self-managed or bespoke managed quote.
SMEs without ops: managed shared or PaaS often beat "cheap" self-managed VPS.
The climax: the most expensive hour is when nobody answers
Calculate RTO in euros, not marketing SLA percentage.
Decide and move forward without blind spots
List current infrastructure tasks and real monthly time spent. Request restore runbook and last test date from managed provider. Compare at least two offers on identical scope. Browse directory and by use case according to ops maturity.
Frequently asked questions
Does managed always include tested backups?
Rarely without reading the contract. Many "managed" offers take automatic snapshots without guaranteed restore or documented RTO — require proof of tested restore.
When does self-managed become cheaper?
When you already have part-time or full-time ops and managed margin exceeds 30–40% of VPS/cloud cost for scope you already master (nginx, Postgres, monitoring).
PaaS = managed?
Partly. PaaS manages runtime and deploy; you remain responsible for code, data, sometimes private network and underlying cloud billing.
How to compare two managed offers?
Same scope: who patches OS? who restores? on-call delay? exclusions (egress, licences)? SLA credits vs real penalties.
Managed or self-managed: decide by multiplying outage hours × business cost — not by comparing two monthly prices.