Sizing Storage for 24/7 Camera Recording

Sizing storage for 24/7 camera recording comes down to one formula: bitrate in megabits per second, times 86,400 seconds a day, divided by 8, gives you gigabytes per camera per day. A 4 MP camera streaming H.265 at 4 Mbit/s writes about 43 GB a day, so eight of them recording continuously for 30 days needs roughly 10 TB. Everything else — resolution, codec, retention — is just changing the inputs to that one calculation.

I’ve watched people guess at this and get burned both ways: buy too little disk and lose the footage they needed, or over-buy 4K cameras and discover the recording drive fills in a week. The math isn’t hard, and doing it before you buy saves you from both. Here’s the formula, a lookup table, and a worked example from my own eight-camera rig.

What’s the Formula for Camera Storage?

Storage per camera per day (in GB) equals the stream bitrate in Mbit/s multiplied by 86,400 seconds, divided by 8 to convert bits to bytes, divided by 1,000 to reach gigabytes. So a camera at 4 Mbit/s is 4 × 86,400 ÷ 8 ÷ 1,000 = 43.2 GB per day. Multiply by your camera count and your retention in days, and you have your total.

The critical insight is that resolution doesn’t appear in the formula — bitrate does. Resolution influences the bitrate you’ll want, but the camera’s configured bitrate is what actually determines file size. A 4K camera throttled to 4 Mbit/s writes the same amount as a 4 MP camera at 4 Mbit/s; it just looks worse doing it, because you’re spreading too few bits over too many pixels. This is why I set bitrate deliberately on every camera rather than trusting the default — it’s the single knob that controls both storage and image quality. The bitrate you can sustain also depends on your camera choices, which I cover in best PoE cameras for a self-hosted NVR.

How Much Storage Per Camera at Common Bitrates?

Here’s the per-camera daily and 30-day storage at the bitrates most home cameras run, assuming continuous recording. Find your camera’s configured bitrate and read across — then multiply by how many cameras you have.

For a whole system, multiply the 30-day column by camera count: four cameras at 4 Mbit/s need about 5.2 TB for 30 days, eight cameras about 10.4 TB, sixteen cameras about 20.7 TB. Those are continuous-recording numbers — the worst case, and the one I plan for, because the two minutes before an event is exactly what motion-triggering tends to miss.

How Much Does H.265 Actually Save Over H.264?

H.265 (HEVC) delivers roughly the same visual quality as H.264 at about half the bitrate, which directly halves your storage. A camera you’d run at 8 Mbit/s on H.264 looks equivalent at around 4 Mbit/s on H.265 — so a fleet that needs 20 TB on H.264 needs closer to 10 TB on H.265. On a multi-camera build recording continuously, that’s the difference between one recording drive and two.

This is why I buy H.265-capable cameras exclusively and confirm my NVR can hardware-decode HEVC before committing. The savings compound across every camera, every day, for the life of the system. The one caveat is decoder support — older or marginal NVR hardware sometimes struggles with H.265, in which case H.264 is the safe fallback at double the storage. The codec and stream details are in RTSP, ONVIF, and the streaming protocols behind IP cameras, and confirming your host decodes HEVC is part of the Frigate setup.

Does Continuous or Motion-Only Recording Change the Math?

Motion-only recording can cut storage by 60 to 80% versus continuous, because most cameras spend most of the day looking at nothing. If a driveway camera only records when Frigate detects a person or car, you’re storing minutes per day instead of 1,440. The catch is that you lose the footage around the event — the approach, the lead-up, the thing that happened just off-trigger.

My compromise: continuous recording on the cameras that matter most (driveway, front door) with shorter retention, and motion-only on the peripheral cameras with longer retention. Frigate makes this easy by letting you set retention separately for continuous recordings and for detection events — keep 7 days of everything but 30 days of clips where a person appeared. That mixed strategy gives me the pre-event context where I need it without paying to store an empty garage 24 hours a day. Sizing the disk for the continuous cameras and treating motion-only as a bonus keeps me from ever running short.

What Kind of Drive Should You Use?

Use surveillance-rated or NAS-rated drives, not desktop drives, because 24/7 recording is a sustained-write workload desktop drives aren’t built for. Surveillance drives (WD Purple, Seagate SkyHawk) are tuned for continuous multi-stream writes and higher workload ratings; NAS drives (WD Red, Seagate IronWolf) suit a recording pool inside a NAS. A desktop drive will work for a while and then fail early under the constant write load.

I keep my recording drive completely separate from the NVR’s OS and config storage, and separate from my Proxmox VM storage, so a full surveillance disk never takes down anything else. Whether a single drive or a redundant pool makes sense depends on how much you’d hate losing footage — I run the recordings on their own pool and back up only the config, since re-recording is automatic but a lost config is a rebuild. For drive reliability data across models, Backblaze’s published drive-failure statistics are the most useful public dataset I know of, and the Frigate documentation covers the retention settings. If you’re building the storage pool on the same box as everything else, keeping it isolated is the same discipline I apply to every service on the home camera network.

Do You Need RAID or Redundancy for Recordings?

For most home surveillance, RAID redundancy on the recording pool is optional, because losing old footage is annoying but not catastrophic the way losing irreplaceable data is. A single large surveillance drive is a perfectly reasonable starting point — if it dies, you lose history, but new recording resumes the moment you swap the drive. I’d rather spend the redundancy budget on more capacity and longer retention than on mirroring footage that’s already low-stakes.

Where redundancy earns its place is uptime, not data safety. If you genuinely need the cameras recording without a gap even during a drive failure, a mirrored or RAID-Z pool keeps recording through the loss of a disk. That’s the choice I made once my camera fleet grew, because a two-day gap while I sourced a replacement drive was a bigger deal than the extra disk cost. The nuance most people miss: RAID is not a backup. It protects against a drive dying, not against a deleted folder or a corrupt filesystem. I back up the NVR config — which is small and painful to recreate — and treat the footage pool itself as expendable capacity, because the recordings regenerate automatically and the config does not.

How Do You Plan for Adding Cameras Later?

Plan storage for the fleet you’ll have in a year, not the one you’re buying today, because camera creep is real and disk is cheapest bought once. When I sized my pool I assumed I’d go from four cameras to eight — which I did — and buying for that target up front saved me a disruptive migration later. A recording pool you have to rebuild because it filled six months in is a weekend you won’t get back.

The practical rule: take your current camera count, add the cameras you can already imagine wanting, run the formula at continuous recording for that number, and add 30% headroom on top. Disk prices per terabyte keep falling, but the hassle of migrating a full recording pool doesn’t, so a little over-provisioning now is almost always the cheaper path. Size once, size generously, and let the retention settings soak up the slack until the fleet grows into it.

The Storage Mistake That Cost Me Footage

Early on I sized my recording drive for the average case and set continuous recording on every camera anyway. The drive filled faster than my mental math said it would, the oldest footage rolled off sooner than my retention target, and the one time I went looking for a clip from three weeks back, it was already overwritten. I’d sized for what I hoped I’d use, not for what continuous recording actually writes.

Now I size for continuous recording on every camera at its real bitrate, add headroom, and treat motion-only savings as breathing room rather than budget. The formula doesn’t lie — a camera at 4 Mbit/s writes 43 GB a day whether or not you wish it wrote less. Do the multiplication honestly, buy the drive that fits the answer, and you’ll never go looking for a clip that already scrolled off the disk.

Keep Building

• Home Camera Network & NVR: The Self-Hosted Guide — where storage fits the whole build.
• Best PoE Cameras for a Self-Hosted NVR — how resolution and bitrate choices drive the disk math.
• Frigate NVR Homelab Setup — setting retention per camera.

What I’d do starting today: pull the configured bitrate off each camera, run the formula for continuous recording over your target retention, add 30% headroom, and buy a surveillance-rated drive that fits. Size for the worst case and the footage you need will always be there.