How to Calculate Storage and Bandwidth for Video Surveillance

The short answer

Everything starts with bitrate — how many megabits per second a camera sends. The rest is arithmetic:

• Archive per day, GB = bitrate (Mbit/s) × 86,400 s ÷ 8 ÷ 1000 = bitrate × 10.8
• Total archive, GB = bitrate × 10.8 × cameras × retention days × recording share
• Network load, Mbit/s = sum of the bitrates of all streams going over a link

Then add a margin for scenes that are busier than expected, filesystem overhead and RAID.

What drives the bitrate

Resolution is only one factor. In practice the bitrate depends on:

• Codec. H.265 usually needs noticeably less bitrate than H.264 for similar picture quality. Smart codecs from camera vendors can reduce it further on static scenes. Make sure the NVR, the viewing software and any integrations support the codec you choose.
• FPS. For most office and retail tasks, 10–15 frames per second is enough. Higher FPS means more data.
• Scene and lighting. Moving trees, crowds, rain and night noise in infrared mode can raise the bitrate significantly.
• CBR or VBR. With a constant bitrate the calculation is exact; with a variable bitrate use the configured maximum for a safe estimate.
• Main stream and substream. The camera sends a high-quality stream for recording and a light one for previews and mobile viewing.

The most reliable source of the number is not a spec sheet but the camera itself: set the resolution, codec and FPS, then look at the real bitrate in the NVR or camera statistics during the day and at night.

Continuous or motion recording

Motion or event recording writes only when something happens. Its effect is captured by the recording share — the fraction of time the camera actually writes. A corridor at night might record rarely; a cash desk records almost all day. If you have no data, measure for a week or calculate for continuous recording and treat the motion scenario as a bonus.

Keep in mind: motion detection can miss events, and pre-record and post-record buffers add time to every event.

Worked examples

Assumptions for clarity (replace them with your measured values): H.265 at 2 Mbit/s per camera, H.264 at 4 Mbit/s, retention in days, plus a 20% margin.

The last two rows show why the codec choice matters at scale: the same retention needs half the disks under these assumptions.

Network load

• 4 cameras × 2 Mbit/s = 8 Mbit/s. Any switch handles this.
• 16 cameras × 2 Mbit/s = 32 Mbit/s of recording, plus substreams if someone watches a grid.
• 32 cameras × 4 Mbit/s = 128 Mbit/s. This exceeds a 100 Mbit/s port, so the NVR uplink and switch-to-switch links must be gigabit.

What to check:

• The NVR's incoming bandwidth limit — every recorder has a maximum for incoming video.
• PoE budget of the switch if cameras are powered over Ethernet.
• Remote viewing goes through the internet upload channel. Use substreams for phones and do not count on watching all main streams remotely.
• Separate VLAN for cameras so video does not compete with office traffic.

Common mistakes

• Counting by resolution only and ignoring night noise and busy scenes.
• Confusing TB and TiB. A disk sold as 4 TB shows up in the OS as less, and RAID takes part of the raw capacity.
• No room for growth. Cameras get added; plan free bays or a larger recorder.
• Using desktop drives. Surveillance-class drives are designed for constant writing.