We run application workloads and their databases on infrastructure we operate ourselves — and we prove the backups restore instead of assuming they do. Small by design: you deal with the person who runs it.
Every database is dumped daily and the newest artifact is restored into a throwaway server automatically. A backup nobody has restored is an untested assumption.
Your database accepts connections from your workload and nothing else. Not a shared network where every tenant can reach every port.
Data sits on dedicated NVMe with a retain-on-delete policy, not on the host's system disk where a rebuild takes it with them.
Staleness, job failure and restore-verification failure each page us. We test that the alerts go red, not just that they exist.
We are not a hyperscaler and we do not pretend to be. There is no availability guarantee, no service credits, and today no off-site copy of your backups. If those matter to you, we will tell you so rather than sell around it — the full list is on the honest limits page, and we would rather you read it before you buy than after something breaks.