On this page
- 01Before you start
- 02NVR, VMS and server
- 03Server roles on a security site
- 04DAS, NAS and SAN
- 05Related storage technologies
- 06Specs that matter for recording platforms
- 07Sizing and choosing the platform
- 08Check that it works
- 09Common problems and how to fix them
- 10Frequently asked questions
- 11Talk to PSS about servers and storage
A security or building site runs on two layers that stay together: servers that process services, and storage that keeps what those services write. Surveillance adds a continuous, write-heavy video stream that must stay for a fixed retention period. Beside it sits site data — configuration, user accounts, event logs, access-control transactions, ANPR plate records, alarm history, and sometimes POS or weighbridge records — small in volume, often kept for years. This guide places the recording decision inside DAS, NAS, SAN and RAID so the video platform matches the rest of the site.
Before you start
- The camera schedule: resolution, codec, frame rate, bitrate and recording mode (continuous, motion or scheduled).
- Retention in days at full frame rate — not a vague “one month”.
- Whether the site is single-building or multi-site, and whether failover is required.
- Which other roles share the rack: file, database, virtualization or backup.
- Room facts: temperature, dust, power circuits and UPS autonomy.
NVR, VMS and server
An NVR, a VMS and a server answer the same question in different ways: where does the data live, and what writes it?
- NVR (Network Video Recorder) — a sealed appliance with embedded firmware, a fixed channel count, built-in PoE ports and two to eight drive bays. Fast to deploy, and free of operating-system patching.
- VMS (Video Management System) — the software layer: live view, playback, user rights, maps, event rules and integrations. It can run inside an NVR or on a server.
- Server — general-purpose hardware (CPU, RAM, network interfaces, RAID controller) hosting the VMS recording service and its storage. You buy it when the appliance model runs out of capacity or features.
Confusing these three is how designs get oversold or undersold. Size storage after the camera schedule is frozen and before the rack, UPS and room cooling are locked — the storage system sets both heat load and rack depth.
Server roles on a security site
A server is built for continuous work: redundant power, enterprise processors, ECC memory, high-speed network ports and remote management. One chassis can carry more than one role:
- File server — shared folders with permissions and protection policies.
- Application server — business applications and processing for users or other systems.
- Database server — structured data for VMS events and access-control transactions.
- Web server — HTTP/HTTPS consoles and portals.
- Virtualization server — several virtual machines on one host.
- Backup server — backup jobs, retention and recovery.
- CCTV recording server — IP camera streams under a VMS, sized for continuous write and contractual retention.
On a campus it is common to see a virtualization host for the VMS, a separate database role for events, a NAS for documents and a dedicated backup target — each sized for its own load.
DAS, NAS and SAN
Enterprise storage is classified by how it connects, how data is accessed and how it is shared.
DAS — Direct Attached Storage
Storage connected directly to one server: internal disks or an enclosure on SATA, SAS or USB. Simple, often cost-effective for a dedicated workload. Sharing across several hosts is limited by design. An NVR or a recording server with local drives is a classic DAS deployment for video.
NAS — Network Attached Storage
A dedicated appliance that gives file-level access over IP, usually through SMB or NFS. Good for departmental shares, document stores, backup repositories and media libraries. Performance depends on the network, the hardware and concurrent load. Traditional NAS is built around files, not low-latency block work.
SAN — Storage Area Network
A dedicated fabric that presents block-level volumes (LUNs) to servers over Fibre Channel or iSCSI. Suited to virtualization clusters, mission-critical databases and a VMS failover pool where several recording hosts must see the same volumes. A SAN does not equal high availability by itself: redundant controllers, paths, power and tested failover must be designed in.
| Feature | DAS | NAS | SAN |
|---|---|---|---|
| Primary access | Directly attached | File-level | Block-level |
| Typical connectivity | SATA, SAS, USB | Ethernet / IP | Fibre Channel or iSCSI |
| Shared access | Limited by design | Multiple clients and servers | Multiple authorized hosts |
| Management | Server-centric | Centralized file services | Centralized storage infrastructure |
| Deployment complexity | Low to moderate | Low to moderate | Moderate to high |
| Typical use | Dedicated servers, NVR / local video | File sharing, backup targets | Virtualization, databases, VMS clusters |
These are general characteristics. Real performance, availability and cost depend on the hardware, network and workload.
Related storage technologies
- Object storage — objects with content, metadata and a unique ID, often reached through S3-compatible APIs. On a surveillance site it usually appears as a cold archive for aged footage.
- Unified storage — file and block on one platform (SMB, NFS, iSCSI and sometimes Fibre Channel).
- JBOD — a collection of disks without RAID protection or striping by itself.
- RAID — disks combined for performance, redundancy, or both. Common levels are listed next.
Common RAID levels:
- RAID 0 — striping; no redundancy.
- RAID 1 — mirroring.
- RAID 5 — striping with distributed parity; minimum three drives.
- RAID 6 — dual parity; minimum four drives.
- RAID 10 — mirrored pairs striped together; typically four drives or more.
RAID is disk organization, not a backup. Accidental deletion, malware, corruption or more failed drives than the array can absorb still destroy data. Keep an independent backup and a tested recovery procedure.
Specs that matter for recording platforms
Sustained write throughput, not capacity alone. A recorder that holds 100 TB but cannot absorb the incoming stream drops frames silently. Size storage twice: once for capacity (TB) and once for throughput (MB/s), then apply the RAID write penalty — roughly four physical operations per logical write on RAID 5, and six on RAID 6.
Drive technology. Use surveillance-rated CMR drives. SMR stalls under continuous writes and does not suit video. SATA is normal for bulk video; SAS where the chassis and controller are enterprise-class; NVMe belongs to the database and hot tier, not to a large archive.
RAID level for video. RAID 5 survives one drive failure but runs unprotected during rebuild. With drives above about 4 TB, a rebuild reads tens of terabytes and the chance of an unrecoverable read error (URE) rises. For large drives, RAID 6 is the default for video; RAID 10 when write performance and fast rebuilds matter more than usable capacity.
Retention is a contractual figure from the client, an insurer or a regulator. Ask for it in days at full frame rate.
Architecture fit. A small single-building NVR is almost always DAS. A campus VMS with failover often lands on SAN volumes or a large DAS enclosure behind the recording hosts. A shared clip-export folder or a backup repository belongs on NAS.
Sizing and choosing the platform
Capacity formula with bitrate in Mbps:
Storage (GB) = Bitrate × 3,600 × 24 × Days × Cameras ÷ 8 ÷ 1,000
It collapses to GB per camera per day = bitrate × 10.8. A 4 Mbps camera therefore writes about 43 GB per day; 30 days of retention is roughly 1.3 TB.
Worked example — 60 cameras averaging 4 Mbps, 30 days of continuous recording:
- Daily write: 60 × 4 × 10.8 = 2,592 GB/day
- Retention: 2,592 × 30 = 77,760 GB ≈ 77.8 TB usable
- Add 15% headroom for motion spikes and databases: ≈ 89 TB
- RAID 6 across 12 drives (≈ 83% usable): raw ≈ 107 TB — for example, 12 × 10 TB
Throughput check: 60 × 4 Mbps = 240 Mbps ≈ 30 MB/s logical; at the RAID 6 write penalty that is about 180 MB/s of physical disk work — comfortable for a modern array, but far beyond two desktop drives.
| Criterion | NVR appliance | VMS server |
|---|---|---|
| Typical channel count | 4–64 channels, fixed | 50–1,000+, licence-based |
| Multi-site / central view | Limited, per-vendor | Native, hierarchical |
| Failover | Rare, model-dependent | Standard (N+1 pool) |
| RAID / drive bays | 2–8 bays, RAID 0/1/5 | 12–60 bays, RAID 5/6/10/60 |
| Storage architecture | Almost always DAS | DAS, SAN or hybrid |
| Integration (access, ANPR, BMS) | Basic, vendor ecosystem | Open SDK / ONVIF / REST |
| Best fit | Single building, ≤ 32 cameras | Campuses, critical sites, long retention |
Move from an NVR appliance to a VMS server when any two of the following are true: more than 64 cameras; more than one site; failover required; retention beyond 30 days with RAID; deep integration with access control, ANPR or BMS; or an IT policy that forbids appliance firmware.
When choosing the wider architecture, weigh performance (throughput, latency, IOPS, concurrent streams), capacity and growth, availability (power, controllers, paths, replication), data protection (backup, RTO, RPO — snapshots and RAID do not replace independent backups), and cost of ownership.
Check that it works
- Capacity and throughput both meet the camera schedule and retention.
- RAID level and drive type match continuous video (CMR, not SMR).
- A second copy of configuration and site data exists off the recording box.
- UPS covers the recorder and the switch that feeds the cameras.
- The rack has rear clearance for a drive swap, and the room has cooling.
Common problems and how to fix them
Retention lasts only a few days
The disks were sized by camera count instead of aggregate bitrate, or quality rose after handover. Recalculate from bitrate × days × cameras, then add capacity or lower quality on views that do not need it.
RAID 5 on large drives failed during rebuild
RAID 5 runs unprotected while rebuilding. On 8–16 TB drives the rebuild is long and URE risk is real. For large video arrays prefer RAID 6, keep a spare on site and replace a failed disk promptly.
SMR or desktop drives in a 24/7 recorder
They stall or wear out under continuous write. Specify surveillance-rated CMR drives from the start.
The recorder survives a power cut; the video does not
The UPS was on the recorder only. Put UPS on the PoE switch that feeds the cameras as well, or the streams stop while the box stays up.
A SAN for a sixteen-camera building
A well-specified DAS NVR would have done the job. A SAN that is underfunded on multipathing and failover costs more and delivers less. Match the architecture to the role.
Continuous video on a general-purpose NAS share
Office traffic and recording compete for the same disk queue. Keep live video on dedicated recording storage; use NAS for exports and backup targets.
The same LUN on two VMS hosts without shared-storage mode
Instant corruption risk. Use a cluster-aware filesystem or the vendor’s shared-storage mode, or keep volumes exclusive to one host.
Frequently asked questions
Is RAID a backup?
No. RAID keeps the array online through a disk failure. A backup is a separate copy you can return to after deletion, corruption, fire or a lost machine. You need both.
DAS, NAS or SAN for cameras?
A single-building NVR is almost always DAS. Shared clip folders and backup repositories fit NAS. A VMS failover cluster that must see the same volumes needs SAN (or a large DAS enclosure behind dedicated hosts). Choose by role, not by catalogue.
How do I get retention in days?
Use GB per camera per day = bitrate × 10.8, multiply by cameras and days, add headroom and RAID overhead. Always confirm the contractual retention in days at full frame rate.
When do I leave an NVR for a VMS server?
When any two of these hold: more than 64 cameras, more than one site, failover required, retention beyond 30 days with RAID, deep integration with access or BMS, or an IT ban on appliance firmware.
Does object storage replace the recorder?
No. It is usually a cold archive tier for aged footage. Live recording still needs sustained write throughput on the primary platform.
Talk to PSS about servers and storage
PSS designs, installs and maintains servers and storage for recordings, access and BMS data, and the site’s applications. Contact the PSS team to book a first site survey.
