CCTV retention math: how many days of footage do you actually get?

The math

A 4MP IP camera recording 24×7 at 15 fps with H.265 encoding produces somewhere between 18 and 25 GB per day, depending on scene complexity. The busier the scene, the more bits the encoder needs.

For a 16-camera system at the lower end of that range:

- 16 cameras × 20 GB = 320 GB/day - 30 days = 9,600 GB raw - RAID 5 with 8× 4 TB drives = 28 TB usable, so 28,000 GB - 28,000 / 320 = 87 days retention, assuming nothing else is on the volume

That's the *upper bound*. In practice, three things eat retention:

1. RAID parity and formatting overhead — RAID 5 over 8 drives loses the equivalent of one drive to parity, plus another 5–10% to filesystem overhead. Real usable is closer to 25 TB. 2. System reserve — the NVR's own OS, VMS database, logs, and metadata. Usually 200–500 GB is gone before recording starts. 3. Other things on the volume — the integrator threw an OS image on the same array, or used a single volume for two NVRs.

That 87 days becomes 65, then 55, then "we're getting 21 days" — which is what shows up at the audit three months later.

The honest retention number

When you size a system, plan for 70% of the rated capacity. The other 30% goes to parity, formatting, metadata, and the things you forgot about.

The 256-camera system needs 173 TB *usable* — which is two racks of disks and a storage array, not a NVR with a couple of shelves.

Continuous vs motion-only

The biggest single lever is recording mode:

- Continuous 24×7 — what the contract usually says, what the system usually isn't doing - Motion-only — 30–50% of the day, 100% of the events. Halves to thirds storage in a typical office or warehouse. Doubles or triples in a quiet corridor after-hours.

For a court complex, the judge's chamber corridor at 2 a.m. should be motion-only. The main entrance should be continuous. Most integrators don't bother setting the schedule per-camera; they record everything continuously and call it a day.

What to specify in the contract

A contract that says "30 days retention" without a definition is the contract you'll regret. Specify:

- Resolution (e.g. 4MP, 2688×1520) - Frame rate (e.g. 15 fps) - Codec (H.265, H.265+, or H.266) - Recording mode per camera class (continuous for entry/exit, motion for internal) - Scene complexity (a busy bazaar is not the same as an empty warehouse) - Storage usable after parity and overhead

When the audit comes and they ask "why are we only at 18 days?", you want the contract to back you up.

Worked example

A 64-camera deployment at GMC Anantnag, recording 24×7 at 4MP / 15 fps / H.265 on a RAID 5 array of 12 × 8 TB drives.

- Per camera per day: ~20 GB - Total per day: 1,280 GB = 1.28 TB - 30 days: 38.4 TB raw - RAID 5 usable: 12 × 8 = 96 TB raw → 88 TB usable after one parity drive - After 30% overhead: 61.6 TB real retention capacity - Real retention: 61.6 / 1.28 = 48 days

So the *minimum* spec for "30 days contracted retention" is 38.4 TB raw, but to actually deliver 30 days you want 60+ TB usable. The difference is what catches most integrators out.

When to call us

If you're sizing a system, send us:

- Number of cameras - Resolution and frame rate per camera - Retention target in days - The site layout (one NVR? distributed? centralised VMS?)

We respond within 4 business hours. The math is part of every proposal we send — the moment we don't, we're not the right integrator.