Infrastructure · Backup
Backup
Creating verified, recoverable copies of data and systems, isolated well enough to survive the event that destroys the original.
Overview
Backup is a recovery capability, not a copy job. The measure of a backup platform is whether a named system can be restored to a defined point within a defined time, proven by test, not whether jobs report success.
Ransomware changed backup design permanently. Attackers deliberately locate and destroy backups before encrypting production, so credential isolation and immutability now matter as much as retention.
How it works
- 01Full, incremental and synthetic full chains balance storage cost against restore speed; snapshots provide fast local recovery points but are not backups on their own.
- 02Application aware processing quiesces databases and mail systems so restores are consistent rather than crash consistent.
- 03Immutability is enforced by object lock, hardened repositories or tape, preventing deletion within the retention window even with administrative credentials.
- 04Recovery objectives drive the design: RPO determines backup frequency, RTO determines the restore method and infrastructure.
Sequence
- 1Production data
- 2Local backup copy (fast restore)
- 3Second copy on different media
- 4Offsite or cloud copy
- 5One immutable or offline copy
- 6Zero errors, verified by restore testing
Reference table
| Tier | Example systems | RPO | RTO | Typical approach |
|---|---|---|---|---|
| Tier 0 | Identity, core network, backup itself | Minutes | < 4 hours | Replication plus immutable copies |
| Tier 1 | ERP, email, primary line of business | 15, 60 minutes | 4, 8 hours | Frequent backups plus warm standby |
| Tier 2 | Departmental applications | 4, 24 hours | 1, 2 days | Nightly backup, restore on demand |
| Tier 3 | Archive, historical data | 24 hours+ | Best effort | Cold storage tiers |
Security considerations
- Isolate backup credentials from production identity; never allow domain administrators to delete backups.
- Require MFA on the backup console and alert on retention or job deletion changes.
- Keep at least one copy immutable or offline, and verify it independently of the primary catalogue.
- Encrypt backups at rest and in transit, and protect the encryption keys separately from the backup system.
Common misconfigurations
- Backup server domain joined with production credentials, encrypted alongside everything else.
- Jobs reporting success while excluding new volumes, databases or SaaS data added since setup.
- No restore testing, so corruption or missing dependencies are discovered during a real incident.
- SaaS data assumed to be backed up by the provider.
IT vs OT
OT backup includes controller programs, HMI projects and engineering configurations, often stored on engineering workstations and rarely covered by enterprise backup. See OT Atlas.
IT Atlas covers Information Technology. Operational Technology is covered separately at otatlas.ai.