Reliability
Backup retention, point-in-time recovery, and read replicas — the three questions that decide how much data a mistake can cost you, and whether you can read from a database without adding load to the one your app writes to. Every figure is sourced from the provider's own documentation; a blank cell means undocumented, not "no."
| Provider | Backup retention | Point-in-time recovery | Read replicas |
|---|---|---|---|
| Aiven | By plan: none on Hobbyist, 2 days on Startup, 14 on Business, 30 on Premium. The magnitude is Startup's 2 days, the entry paid plan. Aiven's high-availability page states 2 days for Hobbyist where the backups page states none — the backups page is the more specific of the two and is what is recorded here. | Bounded by the plan's backup history, so 2 days on the entry paid plan and 30 at the top. Granularity is fine within that: full backups daily with write-ahead logs copied every 5 minutes or per new file, whichever comes first. | Supported |
| Amazon Aurora | 1 to 35 days. Unlike an RDS DB instance it cannot be set to 0 — Aurora backs the cluster volume up continuously and there is no way to turn that off. | Any point inside the backup retention period, so up to 35 days. Aurora's backups are continuous and incremental against the cluster volume rather than periodic snapshots plus log replay, which is why AWS states no performance impact while they are taken. | Supported — 15 per cluster, a quota AWS will not raise |
| Amazon RDS | Between 0 and 35 days, set per DB instance; 0 disables automated backups entirely. The default depends on how the instance was created — one day via the API or CLI, seven days via the console — so there is no single default to quote. Multi-AZ DB clusters cannot be set below 1 day. | Any point inside the configured backup retention period, so up to 35 days. Transaction logs are uploaded to S3 every five minutes, which is what bounds how close to now the latest restorable time can be. | Supported — 15 per primary, raisable on request |
| Crunchy Bridge | 10 days, included in the cluster price rather than sold as a tier, and extendable by configuring additional daily, weekly, monthly and yearly backups to keep. | Any minute inside the 10 days of retained backups. The WAL is streamed every 60 seconds or 16 MB, whichever comes first, and a restore produces a fork of the cluster rather than replacing it. | Supported |
| DreamHost | Daily automated backups on every shared plan. DreamHost does not state how many are kept. | Not documented | Not documented |
| Flywheel | Nightly backups. Flywheel does not state a retention period on the pricing page. | Not documented | Not documented |
| Hetzner | Daily automatic backups with seven slots per server; the oldest is deleted when a new one is taken. Billed at 20% of the server's monthly price. | Not documented | Not documented |
| Hostinger | Weekly backups on the Premium plan; daily on Unlimited. Hostinger does not state how many are kept. | Not documented | Not documented |
| Hostinger VPS | Automatic weekly backups included, plus manual snapshots. Hostinger does not state how many weekly backups are kept. | Not documented | Not documented |
| Neo4j | Daily backups, 7-day retention on the entry-paid (Professional) tier. Business Critical adds 30-day retention with hourly point-in-time restore; Virtual Dedicated Cloud gets hourly backups with 60-day retention. | Not included on the entry-paid (Professional) tier — point-in-time restore is a Business Critical feature, with hourly granularity and 30-day retention. | Supported |
| Neon | 1 day of history on paid plans by default and 6 hours on Free, configurable up to 7 days on Launch or 30 days on Scale. Neon notes that widening the window raises storage cost, so the ceiling is a paid choice rather than an included allowance. | History window of 6 hours up to 1 GB-month on Free, up to 7 days on Launch, up to 30 days on Scale | Supported |
| PlanetScale | Automatic backups every 12 hours on production and development branches, retained 2 days by default. Custom schedules and manual backups can set a longer period, and are charged for it. | Point-in-time recovery through write-ahead log archiving. WAL is kept for as long as the oldest backup, so the default schedule gives a 2-day window and a longer retention extends it. | Supported |
| Pressable | Hourly and daily backups. Pressable does not state a retention period on the pricing page. | Not documented | Not documented |
| Prisma Postgres | Daily backups with 7-day retention on Starter and Pro, 30 days on Business. The free plan has none. | Not documented | Not documented |
| ScaleGrid | Up to five backups per deployment on ScaleGrid's dedicated hosting plans. | Not documented | Not documented |
| Supabase | Daily backups: 7 days on Pro, 14 days on Team, up to 30 days on Enterprise. Free projects get no automatic backups at all — Supabase tells free users to export their own data with the CLI. | Point-in-time recovery is a paid add-on on Pro, Team and Enterprise, and needs at least a Small compute add-on alongside it. Worst-case recovery point is 2 minutes. Without it, restores go back only to a daily backup. | Supported |
| TigerData | Not documented | 3 days on Performance, with cross-region backups available for disaster recovery on higher tiers. | Supported |
| Turso | Not documented | 10 days on Developer. 1 day on the free plan, 30 days on Scaler and 90 days on Pro. | Not documented |