Recording architecture determines where evidence survives, how it can be retrieved and who maintains the system. Compare local, cloud-managed and hybrid designs by tracing the actual data path. Marketing terms alone do not establish storage location, outage behavior or retained quality.
Compare the architectures
| Design | Key planning questions |
|---|---|
| Local NVR | Which cameras and streams are supported, how much usable storage exists, and who maintains the appliance? |
| Server-based VMS | Who owns server sizing, operating-system support, software licensing, redundancy and backups? |
| Edge recording | What happens if the camera or media fails, and can missing recordings be recovered into the main archive? |
| Cloud-managed / hosted | Where are full-resolution video and metadata retained, and what remains accessible without internet or service access? |
| Hybrid | Which copy is authoritative, what is uploaded, and how are gaps and retention differences handled? |
Estimate capacity, then measure
For continuous recording, a useful starting calculation is decimal TB = aggregate Mbps × 0.0108 × days. Sixteen streams averaging 4 Mbps produce about 20.7 TB over thirty days before overhead, redundancy and reserve. A twenty-camera system is not automatically larger than every sixteen-camera system; actual bitrate and recording schedules matter.
Measure representative scenes, including noisy night images, busy movement and weather. Include secondary streams, metadata, incident holds and any required replication. Calculate usable capacity after RAID or other protection. Event recording requires reliable detection and suitable pre/post-event buffers; it should not be sized on an untested assumption that very little happens.
Plan bandwidth beyond recording
Include live viewers, synchronized playback, exports, replication and recovery traffic. A cloud system may queue uploads during an outage and need substantial capacity afterward. Document the upload policy, resolution and retention of each archive. Check whether a local operator can access recordings when external connectivity is unavailable.
Separate retention, redundancy and preservation
Retention describes how long routine recordings remain available. Redundancy addresses selected failures. Incident preservation keeps relevant evidence beyond normal overwrite. None automatically substitutes for the others. RAID is not a backup, and an exported clip may lack context that remains in the original archive. Establish a documented hold and release process.
Verify the operating result
Test the oldest available footage after the required period, not merely on installation day. Confirm timestamps, playback and exports with ordinary roles. Conduct approved outage and restoration tests, including health alerts and any recovery from edge storage. Record who handles disk failure, software updates, subscription changes and eventual service exit.
Compare the proposed configuration with the manufacturer guide and retain the calculations and tests in the planning worksheets.
Related NERSA reference
See NERSA’s cloud, on-premises and hybrid video comparison for the corresponding applied reference.