NVR vs NAS for Surveillance: Which Fits Your Camera Fleet

Leo · 2026-10

An NVR records camera streams and nothing else. A NAS is a general storage server that also records where the vendor firmware supports ONVIF or RTSP ingestion, so the choice between them follows the workload: sites that need file sharing, backup or AI workloads on the same box take the NAS path, and single-job recording sites stay with the NVR.

An NVR (network video recorder: a box whose primary job is ingesting and storing camera feeds) and a NAS (network-attached storage: a general file server on your network) overlap in exactly one task — recording cameras — and diverge in everything else. woCyber, the export brand of a Beijing-based storage hardware manufacturer with mass production in Shenzhen, builds appliances on both sides of that line, and this guide lays out the decision factors from the manufacturing side rather than from a single product sheet.

What Separates an NVR from a NAS in a Camera Project?

An NVR pairs a fixed camera channel count with recording firmware; a NAS pairs flexible drive capacity with general storage services and records cameras only where its firmware explicitly supports it.

The distinction that decides projects is workload count, not the acronym. A recorder that only records is simpler to quote and simpler to replace. A recorder that also serves files, receives office backups and runs containerized applications consolidates three purchases into one — provided the storage architecture keeps camera streams from competing with everything else.

Each row of the table is one job the storage must do, and the appliance that covers all of your rows at the lowest total cost wins. Two measurable inputs sit behind those rows — how many jobs the storage carries, and how much stream load the network holds — and the next two sections take them in order.

Bandwidth, Not Camera Count, Sets the Recording Ceiling

Continuous stream load equals cameras multiplied by bitrate, and that product — not the camera count on the box — decides whether a link, and therefore an appliance, is oversized or undersized.

Put a fleet on the formula (illustrative): 24 cameras streaming 4 Mbps continuously generate 24 × 4 = 96 Mbps. On a 1 GbE link (1,000 Mbps) that is just under 10 percent of capacity, which leaves room for live viewing and backup jobs. A fleet five times larger, at 120 cameras, yields 480 Mbps — nearly half of the same 1 GbE link; a 2.5GbE port brings that ratio back to under 20 percent (480 ÷ 2,500).

• Event-based recording cuts the load to the duty-cycle fraction of the continuous figure: a 16-camera fleet at 4 Mbps with a 20 percent duty cycle averages 16 × 4 × 0.2 = roughly 13 Mbps (illustrative duty cycle).
• Continuous recording scales linearly: double the cameras, double the load, exactly as the 96 and 480 Mbps figures show.
• Once computed stream load plus viewing plus backup traffic approaches a third of the link, the fix is a higher port class or a split across two appliances.

Camera count enters the picture only through this product: bitrate and duty cycle decide the load, not the channel number on the recorder datasheet.

ONVIF and RTSP Keep Camera Vendor Choice Open

ONVIF is the open interoperability standard for IP-based physical security products, and RTSP is the streaming protocol most cameras speak; a recorder that documents both keeps camera vendor choice open for the life of the project.

ONVIF is an open industry forum that provides standardized interfaces for IP-based physical security products, and its profiles are what most camera-recorder pairings rely on. RTSP (Real Time Streaming Protocol) is the application-level protocol defined in IETF RFC 2326 for controlling media delivery with real-time properties, and it is the transport underneath those profiles. A recorder that documents both lets a procurement team replace a failed camera brand without replacing the recording tier.

Power delivery belongs in the same conversation. PoE (Power over Ethernet) carries the network link and the camera’s power over one cable, which removes outlet work at every mount point and keeps the cabling plan independent of the camera brand. The <a href="products/s1-security-kit“>S1 security kit in the woCyber range takes PoE IP cameras over ONVIF with QR-code pairing, so camera choice and cable work stay under the installer’s control.

Edge AI Moves Classification Onto the Recorder

Edge AI moves person, vehicle and pet classification onto the recorder itself, which filters false alerts before footage ever leaves the device and turns the recorder into part of the analytics stack instead of a plain disk target.

Where classification belongs in the requirements, the published S1 materials are one reference point: on-device classification of people, vehicles and pets, a false-alarm figure below 3 percent, retention tiers of 7, 14 and 30 days with event-triggered and continuous loop recording, 100 percent local computation, per-account storage isolation and built-in two-way audio. Whichever appliance a site selects, those items belong in the acceptance test rather than in a brochure conversation.

If AI classification on the recorder is a requirement, put it in the request for quotation as a documented capability with the false-alarm figure stated, then verify it against a live camera during acceptance. Alerts that reach a phone only after a cloud round trip change both the response time and the data path; both belong in the acceptance test.

Capacity Planning Works the Same on Either Appliance

Storage sizing does not differ between the two sides of the NVR vs NAS for surveillance comparison: usable capacity equals raw capacity minus the RAID overhead, and the retention window sets the total.

The constant in that shortcut comes from one day of one megabit: 86,400 seconds ÷ 8 = 10,800 MB, which is 0.0108 TB. Multiply cameras × Mbps × days × 0.0108 and the usable terabytes fall out. For the 30-day case: a 16-camera fleet recording continuously at 4 Mbps needs 16 × 4 × 30 × 0.0108 = about 20.7 TB usable. Four 8 TB drives in RAID 5 yield roughly 24 TB usable (3 × 8 TB), which covers the requirement with margin. The same drives and the same arithmetic apply to a dedicated recorder; the difference is only whether the chassis accepts four drives and grows when the fleet does. The full retention formula and its growth cases are worked through in the dedicated <a href="nas-video-surveillance-retention-storage-sizing“>retention sizing article.

Drive selection adds one constraint worth writing down: surveillance workloads write constantly and read rarely, so the drive class matters as much as the capacity. The <a href="surveillance-hdd-nas-compatibility-guide“>surveillance drive compatibility guide covers how to match drive class to duty cycle, and the answer belongs in the quotation rather than in an assumption.

A Dedicated NVR Fits the Single-Job Site

A site whose storage does exactly one job — recording one building’s cameras — usually spends less on a dedicated NVR than on a NAS, whatever the camera count.

Consider two hypothetical sites. An eight-camera retail unit that records events during opening hours, keeps seven days and shares nothing else fits a compact recorder, and buying extra storage capacity there buys nothing the workload uses. A 28-camera logistics yard that has outgrown its 8-channel recorder, plans two new gates, must hold 30 days of footage and separate per-contractor accounts, and still needs an office file share, consolidates onto the NAS path where one chassis covers recording and file service together. The two hypothetical sites differ in one input only: the number of jobs the storage must do. A camera count alone would have sorted neither site into its column.

For the smaller fleet that still wants camera recording on NAS hardware, the woCyber <a href="products/a2-pro-nas“>A2 Pro is the compact entry on that path: a published 2-bay chassis with RAID 1, a 48 TB maximum and support for ONVIF and PoE cameras.

Four Written Answers Sort the Bids

Four written answers separate qualified bids from enthusiastic ones: camera intake, enforced retention, usable capacity at your drive size, and the paperwork that ships with the goods.

• Camera intake: which ONVIF profiles the firmware supports, whether RTSP streams are accepted, and the PoE budget per port.
• Retention: which durations the firmware enforces and whether the oldest footage overwrites automatically at the configured boundary.
• Capacity: usable terabytes after RAID at the drive size you quote, not the raw chassis maximum.
• Paperwork: which conformity documents arrive with delivery, from the declaration of conformity to test reports.

The <a href="nvr-storage-planning-for-ip-camera-projects“>NVR storage planning guide turns the first two items into a full project checklist, and the arithmetic behind them is already in the sections above.

On the commercial side, the published woCyber terms are a MOQ of 100 units per SKU, rising to 1,000 units where a new PCB respin or new tooling is required, with payment terms confirmed in writing at order confirmation. The published certification set for the product lines covers CE, UKCA, FCC, RoHS, REACH and WEEE, with engineering in Beijing and mass-production manufacturing in Shenzhen.

Storage that holds exactly one job belongs on a dedicated recorder; storage that must also serve files, receive backups or run on-device classification moves the answer to the NAS side of the NVR vs NAS for surveillance comparison, where one chassis carries the jobs together. Two bids that survive the written filter rarely differ on hardware; they differ on documentation, which is what the live-camera acceptance test exists to settle — one camera, one hour, alerts counted against the stated figure.