Simple definition
A backup retains a recoverable version of files, databases or systems so that data can be restored after loss or damage. It is more than a copy stored elsewhere: the appropriate point in time must be available, access rights must be sufficient and the data must be genuinely restorable.
Technical definition
Backup types include full, incremental and differential. A full backup captures the selected dataset. An incremental backup stores changes since a previous backup point; restoration may depend on a chain of points. A differential backup stores changes since the last full backup. Modern products may also use synthetic full backups, changed-block tracking, deduplication, encryption and searchable recovery catalogs.
What is it used for ?
Recover lost or corrupted files, rebuild servers, respond to security incidents and meet business continuity targets. Backups also retain older versions when a retention policy or application requires historical recovery points.
Practical example
A company backs up its VMs nightly to a separate repository and keeps several generations, including a protected copy at a second site. Every month it restores sample files and tests a VM boot. Recovery times are documented so the business can compare results against its continuity requirements.
Common issues
- Backup jobs reported as successful but no restore ever tested.
- Repository using the same physical storage or compromised credentials as production.
- Retention too short to recover from old errors or late-discovered corruption.
- Copies exposed to ransomware with no immutability or account separation.
- Insufficient storage space, bandwidth or backup windows.
Full, incremental and differential backups
A full backup provides a straightforward recovery baseline but can require significant time and storage. Incremental backups reduce daily transferred data by preserving changes, although restoration may depend on several points in a chain. Differential backups can simplify recovery relative to long incremental chains, but their size increases between full backups.
The right method depends on datasets, applications and recovery targets. VM-aware backup products often use changed-block tracking and synthetic backup points. These features help efficiency, but the method's name alone does not prove recoverability.
3-2-1 strategy, RPO and RTO
The 3-2-1 principle recommends three copies in total (production plus two backups), two types of media or technology and at least one off-site copy. The 3-2-1-1-0 extension adds an offline or immutable copy and emphasizes verification without errors.
Recovery Point Objective (RPO) is the maximum acceptable data loss measured in time. Recovery Time Objective (RTO) is the target time to restore the service. A nightly backup may be insufficient for a critical database if business operations require a much shorter RPO.
How to verify that recovery really works
Successful job logs, integrity checks and backup catalogs are helpful but do not replace restoration tests. Recover representative files, databases or whole VMs in an isolated environment. Confirm applications can start and data is consistent, and document how long the process takes.
Backup accounts should be strongly protected and preferably isolated from production administration. Immutable and offline repositories reduce the risk of malicious deletion, but cannot compensate for an inadequate retention policy or incorrectly configured application protection.
Related terms
Key takeaway: backup quality is defined by what can be recovered and how quickly. Extra copies matter only when they are independent, protected and regularly tested.