Enterprise cloud backup: retention, recovery, immutability
Good enterprise backup is about retention periods, demonstrable recovery and copies that attackers or mistakes cannot alter.
Enterprise cloud backup consists of agreed retention periods, demonstrable recovery and copies that cannot be changed or deleted within the retention period (immutability). A backup is only valuable if you can restore it within the required time.
Who is it for: Administrators and security and compliance staff who need to design or assess backups for business-critical data and cloud environments.
What is retention and how do you determine it?
Retention is the period for which a backup is kept. That period comes from several sources: statutory retention obligations, contractual agreements, the time needed to discover errors, and privacy rules that actually limit how long data may be kept.
A common mistake is to retain backups only for a short time. If corruption or an attack is discovered only after weeks, all recent copies are already affected. So choose a tiered schedule with frequent short-term and less frequent long-term retention points. Record how deletion takes place afterwards.
What does immutability mean?
An immutable backup cannot be changed or deleted for a set period, provided it is set up properly (in some modes, administrators with special rights can lift the protection). That helps protect against ransomware, malicious employees and human error.
Check how this is technically enforced and who can lift the setting. A safeguard that can be switched off with a single administrator account offers limited protection. Separate management and separate credentials for the backup environment reduce that risk.
How do you make sure recovery works?
The moment of recovery is not the moment to discover that something is missing. So build recovery into the design.
- Keep a copy at another location or under a different management domain, following the principle of the 3-2-1 rule.
- Also keep the configurations, keys and identity data needed to recover.
- Test the recovery of files, databases and complete systems on a schedule.
- Measure the actual recovery time and compare it with the agreed RTO.
- Encrypt backups and manage keys in such a way that you can still use them in the event of an outage.
How does backup differ from sync and replication?
Sync and replication copy changes immediately, including mistakes and deletions. A backup keeps historical versions. They complement each other for availability, but do not replace each other.
Assessing backup choices
| Topic | Question | Why it matters |
|---|---|---|
| Retention | Which retention periods can you set and enforce | Fits law, contract and the time it takes to detect errors |
| Immutability | Who can lift the protection | Determines resistance to misuse of administrator rights |
| Separation | Is the backup separate from production and identities | Limits the damage if an account is taken over |
| Recovery | How fast and in what form can you restore | Must match the RTO and RPO |
| Manageability | Is there reporting, alerting and proof of execution | Failed jobs must become visible quickly |
Who is it not ideal for?
- Organisations that only use sync and regard that as a backup, because deletions and encryption spread immediately.
- Teams with very low recovery needs, for whom a simpler solution may be enough.
- Situations in which data must be deletable quickly, because long immutability then clashes with privacy obligations; weigh this up in advance.
Frequently asked questions
Does SaaS data also need to be backed up?
Often yes. Vendors protect their platform, but your data, deletions and mistakes generally fall under your responsibility. Check what the service itself offers.
How do you check that a backup has not been compromised?
Use integrity checks, scans and regular restores in an isolated environment, so you know the files are usable and clean.
Who is allowed to restore backups?
Limit this to a small, named group with strong authentication and an approval step, and record every restore in a log.
