In video surveillance systems, the word “backup” is often used for three different things: the RAID array on the recording server, the archive volume old recordings are moved to, and a copy of incident recordings somewhere else. The price of this confusion is paid when a disk array is lost or ransomware encrypts the recording server: recordings thought to be protected turn out not to have been protected at all. In this guide we clarify the difference between live recording, archive and backup, propose a tiered storage model and explain how each kind of recording should be protected.
The difference between recording, archive and backup
| Concept | Purpose | What it protects against | What it doesn’t protect against |
|---|---|---|---|
| Live recording (RAID volume) | Keeping recordings accessible throughout the retention period | Single or double disk failure | Deletion, ransomware, server loss |
| Archive | Moving old recordings to a cheaper, larger tier | Insufficient capacity | Any kind of data loss if the archive is the only copy |
| Backup | An independent, restorable second copy | Deletion, corruption, server and site loss | Ransomware, if the copy is accessible with the same credentials |
| Protected incident recording | Keeping a specific incident beyond the retention period | Automatic overwriting | Deliberate deletion if it’s a single, modifiable copy |
The last column of the table is what really matters. You need a separate protection decision for each tier. RAID isn’t an archive. And an archive isn’t a backup on its own. We covered the protection level of the disk array in the RAID selection guide; the focus here is archive and backup.
Tiered storage model
Most enterprise VMSs can manage recordings across several storage tiers. The typical model is:
- Hot tier (live recording): The recording server’s local or directly attached storage. Write performance is the priority. Retention is kept short; for example, the last few days.
- Warm tier (archive): A larger-capacity volume whose write performance may be lower. The VMS moves recordings older than a set age here with a scheduled job. Some VMSs save space by lowering the frame rate when archiving; this option can affect the evidential value of those recordings, so choose it deliberately.
- Cold tier: Object storage, LTO tape or cloud. Only for incident recordings that must be kept for a long time and footage under legal hold.
Transition times between tiers are derived from the organization’s retention policy, not the other way round. We explained how to set the retention period in the camera recordings and KVKK retention period guide. Footage moved to the archive is also personal data and is deleted when the retention period ends. The archive isn’t a place to postpone the deadline.
Comparison of archive targets
| Target | Suitable use | Retrieval | Immutability |
|---|---|---|---|
| NAS (SMB/NFS) | VMS archive tier | Fast, from within the VMS | Depends on the file system; mostly weak |
| On-premises object storage (S3-compatible) | Incident recordings, long retention | Fast, via API | Strong if the product supports Object Lock |
| Cloud object storage | Off-site copy | Depends on the internet connection and storage class | Strong with Object Lock |
| LTO tape | Large-volume, rarely accessed archive | Slow, manual process | Strong with WORM cartridges or offline storage |
| Cloud VMS (VSaaS) archive | Central access in hybrid architectures | Depends on the service provider | Depends on the service agreement |
Two questions decide the choice. First: does the VMS support this target directly, or is separate backup software needed in between? Second: can a recalled recording be played back inside the VMS with its signature and metadata? An archive that comes back only as files and can’t be re-imported into the VMS costs hours in an incident review. This is also the common gap: the archive exists, but nobody has ever tried to get a recording out of it.
Protecting incident recordings
The vast majority of data a recording system produces is routine footage that will never be watched. The data that really needs protecting is incident recordings, footage under investigation and the system configuration. Targeting these instead of backing up all recordings lowers cost and strengthens protection.
- Marking and locking: At the time of an incident, the operator bookmarks and locks the recording in the VMS (under names such as evidence lock or protect). A locked recording is exempt from automatic overwriting.
- Export: The locked segment is exported in the VMS’s signed or verifiable export format. We explained the integrity mechanisms in the video recording integrity and digital signatures guide.
- Hash value: The SHA-256 hash of the exported files is taken and stored with the incident record.
- Copy to a protected target: The files and hash list are written to an immutable target.
- Documentation: Who locked and exported the recording, when and on what grounds is recorded. The process side is in the incident investigation and chain of custody guide.
The VMS configuration (camera list, recording rules, user and permission definitions, management server database) deserves the same protection. When a recording server is rebuilt without a configuration backup, reconnecting recordings to the VMS can take days even if they’re still on disk. The recordings exist; access doesn’t.
Immutable storage and ransomware
Windows-based recording servers are exposed to the same threats as other servers on the corporate network. An archive share that the recording server can write to and that is accessible with the same domain account gets encrypted along with the recordings during an attack. Options for a protected copy:
- Object Lock: In S3-compatible storage, objects can’t be deleted or overwritten for a set period. In compliance mode, the lock can’t be removed even by an admin account; in governance mode, users with special permissions can bypass the lock.
- WORM tape and offline cartridges: A cartridge removed from the library can’t be accessed over the network.
- Identity separation: The account writing to the archive target is kept separate from the recording server’s domain accounts and has no delete permission.
An example of an incident archive bucket on AWS S3 with a default retention lock (the bucket name, region and period are examples; in this example Object Lock is enabled when the bucket is created and automatically turns on versioning for the bucket. Enabling it later on an existing bucket is also possible, but versioning must be on):
aws s3api create-bucket --bucket example-incident-archive \
--region eu-central-1 \
--create-bucket-configuration LocationConstraint=eu-central-1 \
--object-lock-enabled-for-bucket
aws s3api put-object-lock-configuration --bucket example-incident-archive \
--object-lock-configuration \
'{"ObjectLockEnabled":"Enabled","Rule":{"DefaultRetention":{"Mode":"COMPLIANCE","Days":180}}}'
# Generate a hash list for the exported incident folder and upload
sha256sum incident-2026-0412/* > incident-2026-0412.sha256
aws s3 cp incident-2026-0412/ s3://example-incident-archive/incident-2026-0412/ --recursive
aws s3 cp incident-2026-0412.sha256 s3://example-incident-archive/incident-2026-0412/
Keeping footage containing personal data in the cloud or abroad requires a separate legal assessment. In compliance mode the lock period can’t be shortened; choose a period that doesn’t exceed the upper limit in the retention policy. For general backup principles, the 3-2-1-1-0 backup rule guide is a good starting point for adapting the immutable copy approach to a video archive. If restoring video recordings is part of the facility’s continuity plan, define the test order and retention policy within enterprise backup and disaster recovery.
Restore test
An archive is worth as much as what you can restore from it. An untested archive is an assumption. The test should verify not only that the file exists, but that the recording is usable for its purpose:
- Incident recording: Download a randomly chosen incident folder from the archive, compare the SHA-256 hash values (
sha256sum -c incident-2026-0412.sha256), open the file with the VMS player and verify the signature. - VMS archive: Play back an old recording in the archive tier from the VMS client; check that the timeline, bookmarks and metadata come with it.
- Configuration: Restore the VMS management server backup to an isolated test virtual machine; verify that the camera list, rules and users come back complete.
- Time: Record the restore time in every test. For a tape archive, finding and reading the cartridge counts toward this time too.
Record the test results with the date, data tested, time and outcome. In an audit, this record proves the archive control works.
Archive policy checklist
- Have the concepts of live recording, archive, backup and protected incident recording been defined in writing?
- Are transition times between tiers consistent with the retention policy?
- Is there a procedure for locking, exporting and hashing incident recordings?
- Are the incident archive and VMS configuration backup on an immutable or offline target?
- Is the account writing to the archive target separate from the recording server’s credentials, with limited delete permission?
- Is the deletion of archive recordings whose retention period has ended automatic and provable? The deletion process should be consistent with the camera recording destruction procedure.
- Are restore tests scheduled and their results recorded?
- If a cloud or foreign target is used, has a legal assessment been made?
To design archive tiers and a protected copy for your existing VMS, we can work together under our video recording backup and archiving service; to distribute capacity across tiers, under our recording storage and capacity planning service. The official resource on object lock behavior is the Amazon S3 Object Lock documentation.
Frequently asked questions
Is it necessary to back up all camera recordings?
In most organizations it isn’t, and it at least doubles the storage capacity needed. Continuous recordings are protected by RAID and, if needed, server redundancy. Reserve backup and immutable archive for incident recordings, footage under legal hold and the VMS configuration.
Should archived recordings also be deleted when the retention period ends?
Yes. Footage moved to the archive is also personal data and is deleted at the end of the period set in the retention policy. If you use immutable storage, set the lock period so it doesn’t exceed the retention period; document recordings extended because of an incident separately.
Should object storage or LTO tape be preferred?
Object storage is more practical for frequently recalled incident recordings and automation. LTO tape suits large, rarely accessed volumes kept for a long time and the need for an offline copy. The decision depends on access frequency, retention period and the targets the VMS supports.
If ransomware encrypts the recording server, what is protected?
No copy accessible with the same credentials as the recording server can be considered safe. Protected copies are immutable object storage, offline tape or an archive kept in a separate identity domain. Include the VMS configuration backup in these protected copies too.