Leo · 2026-10
The phrase NVR storage requirements refers to the capacity, bandwidth and parity figures that a retention policy turns into once it meets real cameras. At 2 Mbps across 24 cameras, 30 days of footage lands near 15.6 TB before parity is counted. Four 8 TB drives in a RAID 5 set give 24 TB usable, so the same plan carries 8.4 TB of slack.

Camera counts dominate these conversations, and they are the weakest number in the brief. The bitrate behind each camera moves the total far more than the headcount does.
The storage brief has a second half that never appears in a datasheet: where the recorder physically sits. On EU projects the recorder usually has to sit on your own premises, under your own access rules. An NVR is a recorder that receives and stores camera streams, and placing it in your own server room keeps footage inside your organisation by default, with no upload to a platform nobody in your building administers.
Two lines of a spreadsheet settle most of it, and the work is easier before a supplier conversation than after one.
A 2 Mbps stream writes 21.6 GB per day. Over 30 days, one camera needs 648 GB. Twenty-four cameras need 15,552 GB, which rounds to 15.6 TB of raw video. Motion-triggered recording changes the picture: at a 30 percent duty cycle the same fleet writes roughly 4.7 TB. Two sites with identical camera counts can therefore need very different arrays, so the recording mode belongs in your brief next to the camera list.
Those bitrates are placeholders until you read the stream settings on your own cameras, and re-running the two lines above takes about a minute.
Codec choice moves the bitrate line more than resolution does. Two cameras pointed at the same loading bay can produce different daily volumes depending on the encoder efficiency of each stream, which is another reason the bitrate input should come from your camera’s settings page rather than from a datasheet table.
For multi-camera planning beyond a single retention window, our nvr-storage-planning-for-ip-camera-projects walkthrough covers the project-side checks.
Consider a hypothetical distribution warehouse: 16 fixed cameras on the perimeter and loading bays, 8 more over the sorting floor. The operations manager has been asked to move from 14 to 30 days of retention after two incident investigations ran past the old window. Perimeter cameras record continuously; indoor cameras record on event, because forklift traffic would otherwise fill the disks with empty motion.
The capacity maths: 16 cameras × 648 GB = 10.4 TB of continuous footage. The eight indoor cameras at a 30 percent duty cycle add roughly 1.6 TB. Call it 12 TB raw for the 30-day window.
A second decision sits inside the brief: which cameras justify continuous recording at all. Dropping the eight sorting-floor cameras to event-only was the manager’s trade for keeping 30 days on the perimeter without buying a fifth drive. If the budget forces a choice, retain longer on fewer cameras rather than shorter on all of them; incident timelines rarely respect which camera was cheapest.
Planning against the raw figure alone would be the classic mistake, though. RAID 5 sets one drive aside for parity, and camera firmware updates can bring bitrate changes with them, so headroom has to cover both. On the arithmetic above, a 12 TB workload sits inside four 8 TB drives in RAID 5 with 12 TB to spare, and that slack is the part of the plan that survives a bitrate bump.
The published woCyber line spans three form factors, and they map onto retention workloads differently.
| Unit | Published specification line |
|---|---|
| A2 Pro | 2 bays, 48 TB ceiling, RAID 1, ONVIF / PoE camera support |
| S1 | PoE IP over ONVIF camera input, published 7 / 14 / 30-day retention tiers, other specifications not published |
| X4 | 4 bays, 120 TB ceiling, RAID 0/1/5/6/10, network storage target rather than a camera head |
Where the recorder and the retention policy should arrive as one product, the products/s1-security-kit pairs PoE camera input with the 7 / 14 / 30-day retention options listed on its product page, so the policy in your brief maps to a firmware setting rather than a custom project. Its person, vehicle and pet classification runs on the appliance, with a published false-alarm rate below 3 percent.
Where a camera system already exists and the job is pure storage, the products/x4-nas-server brings four bays, a published 120 TB ceiling and RAID 5 or 6 for parity, which two-bay units cannot offer. Larger estates often mix tiers rather than pick one: a four-bay unit holds the 30-day archive while compact PoE units cover remote gates or outbuildings where cabling a camera back to the core room is impractical. Keep the retention policy identical across tiers so the deletion schedule stays one policy, not three.

Most of what a 30-day retention brief needs is configuration rather than a new product, and that only holds if the details land in writing before the purchase order.
Start with the retention default: which tier leaves the factory, and whether it can be preset on the production line instead of changed on site. Then pin parity to the exact SKU and firmware build, because a family page that mentions RAID 5 says nothing about the unit in the crate. The certification set published for the woCyber product lines is CE, UKCA, FCC, RoHS, REACH and WEEE, which still has to be reconciled against your destination market at quotation.
Commercial terms follow the same rule. Order quantities start at 100 units per SKU, and work that needs new tooling or a fresh PCB respin is quoted from 1,000 units. Warranty is the standard 3-year term with scope confirmed in writing at quotation. Research and development sit in Beijing and mass production in Shenzhen, so a specification change and the production question behind it stay inside one company.
For a first 30-day rollout, hold one retention tier as the default instead of per-camera exceptions, and keep the array well short of full on day one so parity loss and bitrate bumps both fit. For the deletion path, book a demonstration on a sample unit before the production order.
Capacity draws most of the attention in these projects; ingest bandwidth decides whether day one works at all.
Twenty-four cameras at 2 Mbps produce 48 Mbps of continuous ingest, comfortably inside a 1 GbE link. Raise the assumption to 8 Mbps per camera and the aggregate reaches 192 Mbps, which still fits, until playback, export and analytics queries land on the same link at the same time. That concurrency, more than camera count, is what pushes projects toward 2.5GbE. The published X4 specification lists two 2.5GbE ports with link aggregation or failover up to 5 Gbps, while the A2 Pro carries a single gigabit port.
The port-versus-camera priority depends on retrieval behaviour. If staff rarely play back footage and ingest sits far below the link’s capacity, spend on cameras first; the moment playback and analytics share the wire with recording, fund the network port instead. Headroom on the wire absorbs bitrate changes far more cheaply than a capacity forklift later. RAID follows the same logic: a mirrored two-bay array (RAID 1) halves usable capacity by design, RAID 5 on four bays costs one drive, and RAID 6 costs two but survives a second failure during a rebuild, which matters when every drive in the array shares the same age and workload.

Drives are not interchangeable across workloads, and surveillance is the punishing one.
A 24-camera fleet writing continuously puts sequential write load on every spindle for the whole retention window, every day. Drive vendors publish separate duty ratings for models built to run that cycle, and desktop drives usually carry a lighter one. Ask each supplier which drive models they validate on the specific appliance, and keep the answer in the quotation file next to the RAID confirmation. Our surveillance-hdd-nas-compatibility-guide covers the compatibility questions camera integrators run into most often.
Drive class also interacts with rebuild time. Larger drives rebuild for longer after a fault, and a RAID 5 array rebuilding under continuous camera write load spends that window with reduced redundancy. Surveillance-rated drives for the write-heavy arrays, then, plus the supplier’s validated drive list sitting beside the quotation.
Thirty days of retention implies a day 31. Write the deletion behaviour down.
Regulation (EU) 2016/679 expects personal data to be kept no longer than necessary, and surveillance footage counts as personal data the moment a person can be identified in it. This is procurement guidance, not legal advice. What makes the architecture attractive for buyers is structural: footage held on an appliance in your own building, with retention tiers and access logs your team administers, can be evidenced, exported and deleted on schedule. Specifying cameras over ONVIF keeps the recorder swappable too, so the deletion policy does not die with one vendor’s enclosure.
Put three questions to each supplier in writing: which retention tiers the firmware exposes, whether access to footage is logged per account, and what the export path looks like when a formal request arrives. woCyber platforms support on-premises deployment with access audit, and the retention options available on each model are listed on its product page. The compliance-side detail sits in our nas-with-audit-logging-eu-compliance-requirements guide, and the retention-policy angle for EU projects in surveillance-storage-retention-policy-nas-for-eu-projects.
Export is where retention policies get tested. One camera’s full 30-day window at 2 Mbps is 648 GB, and moving it across a 1 GbE link takes a theoretical minimum of about 1.4 hours, illustrative arithmetic rather than a measurement, and longer in practice once the array is also ingesting. Label the estimate as such and plan the export path before a formal request makes it urgent: rehearsing the export on a sample unit turns a formal request into a routine handover.
One caution from the quotation stage: confirm the firmware build number in writing. “Supported” without a build number has a way of meaning “supported somewhere”.

S1 combines an edge AI NVR and private-cloud NAS in one appliance: person, vehicle and pet detection with a false-alarm rate below 3%, 100% local computation and no subscription.
Key takeaways An FCC-recognized lab must hold the right scope for your device's rule part, not just any accreditation. A 2-bay NAS…
November 3, 2026Shared storage for deskside AI clusters refers to a storage appliance on the office LAN that holds datasets, checkpoints and embeddings for…
November 3, 2026A cloud storage subscription cost is the recurring fee a business pays to keep its files on a provider's servers, billed per…
Ask our export team for the OEM/ODM cooperation guide and a written quotation.