Surveillance footage and broadcast masters look identical on disk and behave nothing alike.

The design consequence is that surveillance archives are optimised for cost per written terabyte and for provable deletion, while media archives are optimised for findability and restore speed. A single platform can serve both, but only if the policies applied to each are genuinely separate.

The figures here are constructed to demonstrate the arithmetic. They are not measurements from any site, and real bitrates vary with codec, scene motion and compression settings.

Take 400 cameras at 4 Mbps, recording 24 hours a day.

• 4 Mbps is 0.5 MB per second, so one camera writes 0.5 × 3,600 = 1.8 GB per hour
• Over 24 hours: 1.8 × 24 = 43.2 GB per camera per day
• Across the estate: 43.2 GB × 400 = 17.28 TB per day

Now apply retention windows to the same estate.

The point is the ratio. Doubling the estate to 800 cameras at 30 days gives about 1,037 TB. Leaving the estate at 400 cameras and extending retention to 90 days gives about 1,555 TB, half as much again, from a policy decision that costs nothing to make in a meeting. Going from 30 days to a year multiplies the requirement by roughly twelve while the camera count never changes.

So retention is the variable to interrogate first. Two questions usually shrink the number: does every camera need the same window, and does every camera need full frame rate for the whole window. Tiering by camera class, and stepping down to a lower-bitrate copy after the first fortnight, moves the total far more than any storage discount will.

How long footage must be kept is not something we can tell you. Periods and their interpretation are a legal and policy question for your own advisers, and they differ by sector, by site and by what the footage shows.

The recording tier is a VMS problem. The archive tier begins where the VMS stops, and it has three jobs.

• Absorb continuous ingest without back-pressure. If the archive stalls, the VMS retention window silently shortens. Monitoring should alarm on archive write lag, not only on disk capacity.
• Hold cheaply for the long tail. Footage older than a month is almost never read. Removable or optical media suits it, provided the catalogue stays online so a request can be located before anything is mounted.
• Delete on schedule and prove it. Deletion is a feature here, not a risk. The archive should record what expired, when, and under which policy.

Onboard and remote estates add a wrinkle: vehicles and remote sites accumulate footage locally and offload in bursts when they reach a depot or a link. The archive has to cope with a day's worth of many cameras arriving in minutes.

QStar and DISC publish a deployment at Nottingham City Transport covering onboard vehicle video retention. That is a manufacturer's reference, not an Aban Smart customer, and we present it as an illustration of the pattern rather than as our own work.

Production archives fail differently. The file is there, the tape or disc is readable, and no one can tell what is on it. An archive without a catalogue is a landfill with an index card.

This is why the archive belongs behind the asset management system rather than beside it. QStar documents an integration with Iconik by Backlight in which assets are archived through the Iconik Storage Gateway and retrieved through QStar Archive Manager, while remaining visible and searchable in Iconik. Editors keep browsing proxies in the interface they already use, and the high-resolution master moves off expensive production storage to a lower-cost tier as a project closes.

QStar's archive software presents these tiers through standard file system and S3-compatible interfaces, so the underlying media can be disk, object storage, cloud or tape without the MAM needing to know. Three practical notes:

• Keep proxies on fast storage permanently. They are small, and they are what people search.
• Archive at project close, not at some age threshold. Age is a poor proxy for "finished".
• Store the metadata with the essence. A MAM database that dies without the archive takes the archive's meaning with it.

For the media layer itself, see optical archive systems and the wider immutable and WORM storage picture.

The moment footage might be used in a proceeding, the requirement changes from storage to evidence, and evidence is about proving the file was not altered between the camera and the courtroom.

Four things carry that argument.

1. A hash captured at ingest. Compute a cryptographic checksum when the footage lands in the archive, store it separately from the file, and re-verify on every export. An exported clip with a matching hash is a much stronger artefact than an exported clip alone.
2. Export as a copy, never as a move. The archived original stays untouched under its retention lock. What leaves is a derived copy with its own record.
3. A custody log. Who requested the export, under what reference, who authorised it, what was exported, and where it went. For removable media, the same log tracks the physical item.
4. Write-once media where the stakes justify it. Retention-locked appliances resist deletion by policy. Optical WORM resists it at the recording layer, which is not rewritable, so an administrator with full rights has no data path for altering what was written. DISC documentation cites 50-plus-year Blu-ray media life with true WORM support, and QStar Archive Manager can virtualise a DISC appliance so it appears as a CIFS or NFS network share. Note that some of that DISC material dates from an earlier product generation, and capacity figures should be confirmed against current documentation rather than taken from a brochure.

No storage product makes footage admissible. Admissibility depends on procedure, authorisation and legal interpretation, and it is a question for your own advisers. What the storage layer can do is remove the argument that the file changed.

The retention-and-evidence pattern here overlaps with government estates, and the sheer-count problem overlaps with telecommunications. Browse the full set of industries, or request an assessment with your camera count, bitrate and current retention window, which is enough to start.