NVR Storage and Bandwidth Sizing Guide for Commercial Camera Systems
The most expensive phone call in surveillance is the one where the customer asks for footage from three weeks ago and the recorder only kept nine days.
Retention failures are almost never hardware failures. They are math failures made at quote time: a bitrate guessed too low, a frame rate left at the default, or a retention requirement nobody wrote down. This guide shows the math so you can size NVR recorders and VMS servers with numbers instead of hope.
Step 1: Get the retention requirement in writing
Size storage backwards from retention, never forwards from a drive you happen to stock.
- 30 days is a common commercial baseline.
- Regulated sites (cannabis, gaming, some healthcare and financial) often have longer, written requirements. Get the actual rule from the customer.
- Insurance and HR often drive retention more than security does. Ask who requests footage and how late they usually ask.
Write the number on the quote. If retention is ever disputed, that line protects you.
Step 2: Estimate bitrate per camera
Bitrate is the single input that decides everything else. It depends on resolution, frame rate, codec, scene activity, and night-time noise.
Typical continuous-recording starting points (validate with the manufacturer calculator and a live site check):
| Resolution | Frame rate | H.264 | H.265 | H.265 + smart codec |
|---|---|---|---|---|
| 2MP (1080p) | 15 fps | 4 Mbps | 2 Mbps | 1–1.5 Mbps |
| 4MP | 15 fps | 6 Mbps | 3–4 Mbps | 2–2.5 Mbps |
| 8MP (4K) | 15 fps | 12 Mbps | 6–8 Mbps | 3–5 Mbps |
| 8MP (4K) | 30 fps | 20 Mbps | 10–12 Mbps | 6–8 Mbps |
Three things push real bitrate above these numbers:
- Busy scenes. Trees, traffic, rain, and crowds generate more data than an empty hallway.
- Night. Low light adds sensor noise, and noise compresses poorly. Night bitrate can exceed day bitrate.
- Smart codec settings left off. Many cameras ship with dynamic GOP and smart codec disabled.
When in doubt, size to the higher number. Under-sizing costs retention; over-sizing costs a drive.
Step 3: The storage formula
Storage (TB) = Bitrate (Mbps) × 86,400 × Days × Cameras ÷ 8,000,000
Where 86,400 is seconds per day and dividing by 8,000,000 converts megabits to terabytes.
Worked example: 16 cameras at 4 Mbps, 30 days.
4 × 86,400 × 30 × 16 ÷ 8,000,000 = 20.7 TB
Then add headroom:
- About 10% for file system overhead and the gap between marketed and formatted drive capacity.
- RAID overhead if the server uses RAID 5 or 6 (one or two drives of capacity lost to parity).
That 20.7 TB job realistically wants 24 TB usable.
Quick-reference retention table
Continuous recording, 4 Mbps per camera (a typical 4MP H.265 stream), before overhead:
| Cameras | 15 days | 30 days | 60 days |
|---|---|---|---|
| 4 | 2.6 TB | 5.2 TB | 10.4 TB |
| 8 | 5.2 TB | 10.4 TB | 20.7 TB |
| 16 | 10.4 TB | 20.7 TB | 41.5 TB |
| 32 | 20.7 TB | 41.5 TB | 82.9 TB |
For 4K cameras at about 8 Mbps, double every figure.
Step 4: Continuous vs. event recording
Motion or AI-event recording can stretch retention dramatically, because the recorder only writes when something happens.
- Event-only recording on a quiet site can cut storage by 50–80%.
- The trade-off: you lose what happened between events. For investigations, a missing lead-up is often the part that matters.
- A common compromise: continuous recording at a reduced substream bitrate, plus full-resolution recording on AI human and vehicle events.
Never size storage on event recording alone unless the customer accepts the gaps in writing.
Step 5: Check the network, not just the disks
A recorder that has room for the video still fails if the video cannot reach it.
- Total ingest: 32 cameras × 8 Mbps = 256 Mbps continuous. Fine on gigabit, but leave room for playback and export.
- Playback and remote viewing add load on top of recording. Plan for at least two times sustained camera bitrate on the uplink to the recorder or server.
- Switch uplinks: larger sites should use a 10G uplink between the camera switch and the server. See our PoE switch buying guide for uplink and VLAN planning.
- Recorder throughput specs: every NVR lists a maximum incoming bandwidth. Total camera bitrate must stay under it, with headroom.
Step 6: Choose NVR or server
| Situation | Better fit |
|---|---|
| Up to about 16 cameras, single site, simple retention | PoE NVR recorder |
| 32+ cameras, long retention, or RAID required | Dedicated VMS server |
| Multiple operator workstations or heavy playback | VMS server (decode load matters) |
| Analytics and forensic search across many cameras | VMS server |
If the job is outgrowing a recorder, read why integrators build custom Wisenet WAVE servers.
Commissioning checklist
Before you leave site:
- Confirm the recording schedule and stream settings on every camera.
- Record the actual bitrate per camera from the recorder, both day and night if possible.
- Recalculate retention from the measured numbers, not the quote numbers.
- Check drive health status and that every drive is recording.
- Set disk-full behavior to overwrite oldest, not stop recording.
- Give the customer the retention figure in writing.
Every camera and NVR recorder NorthStar stocks is NDAA compliant and backed by a 3-year warranty. Send us the camera list and retention requirement and we will return a recorder or server configuration with the storage math shown.
Frequently asked questions
How much storage do I need for 30 days of security camera footage?
Use Storage (TB) = bitrate in Mbps x 86,400 x days x cameras / 8,000,000. Sixteen cameras at 4 Mbps for 30 days needs about 20.7 TB before overhead. Add roughly 10 percent for formatting and file system overhead, plus RAID parity if used.
What bitrate does a 4K security camera use?
A typical 8MP (4K) camera at 15 fps uses around 12 Mbps in H.264, 6 to 8 Mbps in H.265, and 3 to 5 Mbps with a smart codec. Busy scenes and night-time noise push bitrate higher, so validate on site.
Does H.265 really cut storage in half?
Roughly. H.265 commonly reduces bitrate by 40 to 50 percent versus H.264 at similar quality, and smart codec features such as dynamic GOP can reduce it further on quiet scenes.
Should I use motion recording to save storage?
Event recording can cut storage by 50 to 80 percent on quiet sites, but it loses the footage between events. A common compromise is continuous low-bitrate substream recording plus full-resolution recording on AI human and vehicle events.
When should I use a VMS server instead of an NVR?
Consider a dedicated VMS server above roughly 16 to 32 cameras, when long retention or RAID is required, or when several operators view and play back video at the same time.