The Snapshot Frequency Matrix for Recovery

Snapshot frequency drives RPO. The matrix that picks the right cadence per workload class.

RPO drives the cadence

Recovery Point Objective is the maximum acceptable data loss. Per service, in business time. RPO of 1 hour means the team accepts losing up to 1 hour of data in a worst-case recovery.

Snapshot frequency directly bounds RPO. Daily snapshots: RPO of 24 hours. Hourly: 1 hour. Per-15-minutes: 15 minutes plus reconstruction time.

Below per-15-minute snapshots, continuous replication is usually cheaper and tighter. Snapshots at high frequency become operationally expensive.

The matrix by criticality

Critical (financial transactions, user data): continuous replication, plus daily snapshots for archive. RPO seconds.

Important (production databases, core configurations): hourly snapshots, plus continuous replication for primary. RPO 1 hour.

Standard (reporting databases, analytics): daily snapshots; replication optional. RPO 24 hours.

Low-criticality (logs, derived data): daily or weekly. Often regenerable; recovery from source rather than snapshot.

Retention policy per tier

Critical: 30 days hot, 90 days warm, 7 years cold for compliance. Cost is real but justified.

Important: 14 days hot, 30 days warm. Most recovery scenarios fit within this window.

Standard: 7 days hot, 30 days warm. Beyond that, data is rarely needed.

Low-criticality: 7 days. Cheap to keep; cheap to lose.

Test recovery, not just snapshots

Snapshot existence is not the same as recoverability. Quarterly: pick a snapshot, restore to a clean environment, verify integrity.

Time the recovery. RTO (Recovery Time Objective) compounds with RPO. A 1-hour RPO with 8-hour RTO leaves 9 hours of customer impact.

Document the procedure. Untested procedures fail under pressure. The first restore in production should not be the first restore ever.

Cost considerations

Snapshot storage scales with frequency × retention × data size. A 1TB database with hourly snapshots and 30-day retention is 720 snapshots, costing real money.

Incremental snapshots help. Most cloud providers store only changed blocks; the marginal cost per snapshot is small after the first.

Cross-region replicated snapshots double the cost but provide region-failure protection. Required for compliance in many industries.