Recording, storage, and retention

Check recording from capture through storage, indexing, retrieval and playback. Compare the retained footage with the required duration and quality.

Overview#

Recording is a pipeline, not a disk allocation. A system can retain files yet fail to provide complete, correctly timed, decodable, searchable, authorised evidence.

Recording lifecycle#

capture -> encode -> receive/buffer -> segment -> index -> replicate/protect
 -> search/playback -> export/hold -> expiry/deletion -> media retirement

Track media segment, recording track, camera/source, codec/configuration, start/end source and receive times, continuity/gaps, integrity/provenance, storage tier, retention class, hold state, and deletion outcome.

Recording policy#

Define per purpose/source:

• continuous, event triggered, schedule, manual, or failover recording;
• pre/post event buffer and trigger source;
• primary/substream, audio, metadata, overlays, and privacy masking;
• frame rate, resolution, bitrate/quality, keyframe interval, and codec;
• edge, site, cloud, redundant copy, and failure behaviour;
• retention start point and deletion semantics;
• user/role rights for view, export, protect/hold, and delete.

Event triggered recording depends on detector availability, time, rule/configuration, prebuffer, and recovery; don't advertise its nominal retention as continuous coverage.

Capacity model#

Estimate from measured/contracted peak and distribution, not only average bitrate:

ingest = sum(stream peak bitrate × duty cycle) + protocol/index overhead
usable capacity = retention × ingest × resilience/rebuild/headroom factors

Account for variable bitrate scene changes, keyframe bursts, audio/metadata, multiple profiles, failover/catch up, filesystem/object overhead, replicas/erasure coding, snapshots, export staging, audit, rebuild load, and growth. Thermal, power, network, controller, IOPS, and object/API limits can fail before raw capacity.

Edge and gap reconciliation#

ONVIF Profile G provides a standardised scope for edge recording configuration/control/retrieval. Actual gap fill sequencing, bandwidth throttling, retention competition, clock correction, duplicate segments, and device support require product verification.

After outage, preserve source sequence/time and mark unrecoverable intervals. Catch up traffic must not starve live recording. A later recovered segment may overlap central recording; deduplicate without discarding provenance.

Retention and deletion#

Retention is a policy decision tied to purpose, risk, law, contract, investigation holds, and storage cost. “30 days” must define whether age is capture time, ingest time, or calendar policy; how unavailable cameras and daylight saving transitions appear; and how protected incidents override expiry.

Deletion must cover primary, replicas, caches, exports, thumbnails, analytics indexes, backups, cloud lifecycle/versioning, and decommissioned disks, while preserving authorised holds. Record deletion request, authority, policy, target scope, completion/evidence, and failures. Cryptographic erasure only works when keys and copies are correctly scoped.

Resilience and evidence#

• Separate recording continuity from configuration/database backup.
• Check backup completion, then test restoration, search and playback.
• Monitor gap duration, late writes, corrupt/undecodable segments, time drift, index lag, capacity forecast, disk/object errors, replica state, and retention failures.
• Restrict direct storage access; use service identities and encryption with managed keys.
• Preserve native master evidence and transformation history for exports.

Return to Video surveillance systems.