How much storage do security cameras need?

Resolution is an input to the encoder—not the storage equation

Two cameras can output the same resolution and frame rate while producing very different bitrates. A quiet indoor corridor is easier to compress than rain, foliage, traffic, sensor noise, or a busy retail floor. Constant bitrate settings cap or target traffic differently from variable bitrate settings. Quality targets, GOP structure, audio, analytics metadata, and firmware all matter.

That is why a storage worksheet should accept the stream bitrate directly. Before installation, use the exact vendor planning value for the intended profile. After installation, observe typical and busy-period traffic and revise the capacity scenario.

Work from recorded seconds

For one camera, multiply megabits per second by the seconds actually recorded. Continuous recording uses 86,400 seconds per day. A 12-hour schedule uses half that. Event recording reduces time only when the recorder truly omits non-event footage; pre-roll and post-roll make naive motion percentages optimistic.

• Raw decimal TB = cameras × Mb/s × days × hours/day × 3,600 × activity ÷ 8 ÷ 1,000,000.
• Add container/filesystem overhead explicitly.
• Then map the target to RAID-usable capacity and subtract working reserve.
• Finally compare estimated peak traffic with recorder and network limits.

A worked eight-camera example

Eight cameras averaging 4 Mb/s create 32 Mb/s of recording traffic. Over 30 days of continuous recording, that is about 10.37 decimal TB of raw video. A visible 10% container/filesystem allowance raises the recording target to about 11.40 TB.

That is not yet a drive shopping list. Four 4 TB drives in RAID 5 offer a simple 12 TB before vendor overhead and reserve—too tight if the plan also keeps 15% free. A larger layout, lower verified bitrate, shorter retention, or different policy is needed. Keeping those steps separate exposes the trade instead of hiding it.

Use scenarios when the site is unknown

A single precise-looking answer is false confidence before commissioning. Build a low case for quiet conditions, a base case from representative vendor data, and a high case for difficult night or high-motion periods. If all three fit, the decision is robust. If the high case breaks the design, measure the installed stream before promising retention.

Retention is an operational result, not a one-time calculation. Recorder configuration can change after a firmware update, camera replacement, quality adjustment, or analytics rollout. Recheck it when the system changes.