Lead Forensics

When does a CCTV deployment need petabyte-scale SAN storage?

A CCTV deployment needs petabyte-scale SAN storage when sustained video ingest, long retention periods and the number of recording servers push total capacity into the high hundreds of terabytes or beyond, and local server disks can no longer be managed or protected efficiently. In practice, that means large camera estates, high bitrates and retention of 30 days or more.

The deciding factors are capacity, aggregate write throughput and operational risk, rather than camera count alone. The sections below explain how to calculate requirements, which sectors reach this scale, and how to plan and validate a SAN for surveillance servers.

How do you calculate CCTV storage requirements at scale?

CCTV storage is calculated by multiplying each camera’s average bitrate by its recording time, then summing across all cameras and applying the retention period plus overhead. The core formula is bitrate (Mbps) ÷ 8 × 86,400 seconds per day × retention days, converted to terabytes, before adding RAID parity, spares and growth headroom.

As a purely hypothetical illustration, 1,000 cameras each averaging 4 Mbps equal 4,000 Mbps in aggregate, or roughly 500 MB/s of continuous write traffic. Over 30 days that is approximately 1.3 PB of raw recorded video before any RAID overhead. These figures are an example of the method, not a guaranteed result for any real system.

The inputs that move the result most are:

  • Resolution and codec: H.264 and H.265 produce very different bitrates for the same scene, and smart codecs vary the stream by scene activity.
  • Frame rate and recording mode: continuous recording at 25 fps stores far more than motion-triggered recording.
  • Scene complexity: busy concourses, low light and noise raise bitrates substantially compared with static corridors.
  • Retention policy: legal, insurance or regulatory requirements often set the retention period rather than operational preference.
  • Metadata and analytics: AI analytics generate metadata and sometimes secondary streams that add storage and change the I/O pattern.

Add margin for RAID parity, hot spares, file system overhead and future camera additions. Calculations built on vendor maximum bitrates tend to overstate need, while those built on idealised averages tend to understate it, so validated figures matter. Secure Logiq’s guide on selecting a server covers the sizing logic in more detail.

Which deployment sizes and sectors typically reach petabyte scale?

Petabyte-scale CCTV storage is typically reached in city-wide surveillance, transport hubs, large stadiums and venues, universities, critical infrastructure sites and major retail or leisure estates, usually where there are well over a thousand cameras, high-resolution streams or long retention requirements.

Scale comes from combinations rather than a single threshold. Sectors that commonly approach it include:

  • Public spaces and smart cities: thousands of cameras, often with ANPR and analytics, and retention driven by evidential needs.
  • Transport hubs: airports, rail stations, seaports and mass transit networks combine high camera density with demanding analytics.
  • Stadiums and arenas: high frame rates, dense coverage and live event safety requirements.
  • Education campuses: multi-site estates that consolidate CCTV with other security subsystems.
  • Utilities and critical infrastructure: expansive sites with perimeter detection, where recording continuity is critical.
  • Shopping centres and leisure: continuous multi-stream HD recording combined with retail analytics.

A smaller site can also reach this scale through 4K or multi-sensor cameras recorded continuously over extended retention periods. Camera count is a useful indicator, but bitrate and retention determine the final figure.

When is direct-attached storage no longer enough?

Direct-attached storage (DAS) is no longer enough when storage must be shared, expanded or protected beyond what individual recording servers can provide, typically when capacity per server hits practical limits, many servers need pooled capacity, or failure of one server would create unacceptable recording gaps.

DAS works well where each recording server owns its disks and the system is modest in size. It becomes a constraint in a number of situations:

  • Stranded capacity: some servers fill while others sit half empty, because capacity cannot be reallocated.
  • Management overhead: dozens of servers with separate arrays multiply monitoring, maintenance and drive replacement tasks.
  • Failover limitations: if a recording server fails, its locally attached video can be unavailable until it is repaired, unless failover is designed around shared storage or replication.
  • Expansion limits: chassis drive bays cap how far a single server can scale without adding more servers.
  • Centralised retention policy: a single pool is simpler to size and audit than many separate arrays.

DAS remains valid for many projects, and a well-built server with hardware RAID and hot spares can store a very large amount of video. A SAN is justified by operational and resilience requirements, not by capacity alone.

What’s the difference between SAN, NAS, and server-based storage for CCTV?

A SAN provides block-level storage over a dedicated network to multiple servers, NAS provides file-level storage over a standard network, and server-based storage keeps disks inside or directly attached to the recording server. Each trades simplicity, scalability and performance differently.

Server-based (direct-attached) storage

Disks sit in the recording server or an attached enclosure. This is the simplest and often the most cost-effective design, with predictable write performance because no storage network is involved. Its limits are scale and flexibility, as described above.

NAS

NAS shares files over Ethernet using protocols such as SMB or NFS. It is easy to deploy and suits archive or secondary storage, but file-level protocols and shared network paths can introduce latency and contention under heavy continuous video writes. Suitability depends on the VMS, which should be checked against the vendor’s storage guidance.

SAN

A SAN presents block storage to recording servers over technologies such as Fibre Channel or iSCSI, which servers treat like local disks. It allows pooled capacity, shared failover and centralised management at scale. It also adds cost, network design complexity and a shared dependency that must be engineered for resilience.

No option is universally best. SAN is the right choice when pooling, shared failover and large-scale management outweigh the added complexity. Which protocols and features a particular platform supports should be confirmed against official product documentation, not assumed.

What resilience and performance features should a surveillance SAN have?

A surveillance SAN should combine redundant components, protected RAID, sustained sequential write performance and clear failure behaviour, so that recording continues through a disk, controller, path or power fault. It should be sized for continuous video ingest rather than the random I/O workloads typical of general IT.

Key features to specify and evaluate include:

  • Protected RAID with hot spares: RAID 5, 6 or 60 with hot-swappable spares, so a failed drive can be replaced and the array can rebuild without interrupting recording. RAID is protection against disk failure, not a backup.
  • Redundant controllers and paths: dual controllers and multipath connectivity remove single points of failure between recording servers and storage.
  • Redundant power: dual hot-swappable power supplies on storage and servers.
  • Sustained write performance: throughput should be validated under continuous load, including during a rebuild, since degraded-mode performance is where many designs fail.
  • Surveillance-rated drives: enterprise drives designed for 24/7 write-heavy operation.
  • Monitoring and alerting: proactive health reporting for drives, controllers and capacity.
  • Metadata handling: in analytics-heavy systems, separating video from metadata can reduce fragmentation and improve archive stability.

Redundancy does not automatically guarantee uninterrupted recording, so failover behaviour should be tested rather than assumed. Secure Logiq’s Logiqal Core Architecture and Logiqal Video Volume address video-optimised compute and software-defined storage for large-scale surveillance, including automated failover and failback.

How should petabyte-scale CCTV storage be planned and validated?

Petabyte-scale CCTV storage should be planned from measured or simulated camera workloads, designed around the full recording chain from cameras to network to servers to storage, and validated by testing sustained throughput and failure scenarios before deployment, not by relying on specification sheets.

A practical approach follows these steps:

  1. Define requirements: camera counts, resolutions, codecs, frame rates, recording modes, retention periods and analytics workloads.
  2. Calculate bandwidth and capacity: document assumptions so the figures can be audited, and include growth headroom.
  3. Design for the whole system: storage performance depends on network capacity, recording server load and VMS behaviour, not just drive specifications.
  4. Plan resilience and recovery: specify failover, rebuild behaviour and, where required, a separate backup or replication strategy.
  5. Simulate before purchase: Secure Logiq’s Logiqal Benchmark utility can model multi-camera networks, including resolutions, compression formats, frame rates and analytics workloads, to test server and storage throughput limits before physical deployment.
  6. Test failure and degraded modes: pull a drive, fail a path and confirm recording continues and rebuilds do not cause dropped frames.

Real deployments illustrate the approach. Secure Logiq’s Qatar University SAN storage case study shows how large-scale storage was applied in an education environment, and its system design service provides storage calculations, schematics and documentation for specifications and tenders. Where standard configurations do not fit, bespoke solutions can be engineered around the project’s requirements.

If you are sizing a surveillance estate approaching petabyte scale, contact the Secure Logiq team to discuss your retention, throughput and resilience requirements before committing to an architecture.

Frequently Asked Questions

How do I get started sizing a SAN if I only have a rough camera list?

Start by grouping cameras by type (resolution, codec, frame rate, recording mode) and assign each group a realistic average bitrate, ideally from test recordings of your own scenes rather than datasheet maximums. Sum the bitrates for aggregate write throughput, multiply by retention days for raw capacity, then add RAID parity, hot spares, file system overhead and 20-30% growth headroom. Document every assumption so you can refine the figures as real camera data comes in.

What happens to recording if a SAN controller or storage path fails?

With dual controllers and properly configured multipath connectivity, recording should continue over the surviving path, though performance may be reduced. This only works if the failover has been designed and tested end to end, including the recording servers’ multipath drivers and the VMS behaviour. Always simulate a controller and path failure during commissioning and confirm there are no gaps or dropped frames.

Is a RAID array enough, or do I also need a backup for CCTV footage?

RAID protects against disk failure, not against deletion, corruption, controller faults, fire or theft. Whether you need a separate backup or replication depends on your evidential and regulatory requirements. Many large estates replicate critical footage or exported incidents to a second array or site, even if routine footage is not backed up in full.

Why does performance sometimes drop during a RAID rebuild, and how can I plan for it?

During a rebuild, the array must read from all remaining drives to reconstruct data while still handling continuous video writes, which can cut available throughput significantly. To plan for this, choose RAID levels with sufficient parity (such as RAID 6 or 60), keep hot spares in place, and leave performance headroom above your calculated ingest rate. Then test sustained writes in a degraded state before go-live rather than relying on specification sheets.

What are the most common mistakes when specifying storage for large CCTV systems?

Common mistakes include sizing from vendor maximum bitrates or idealised averages, ignoring metadata and analytics streams, forgetting growth headroom, and focusing on capacity while neglecting sustained write throughput. Others are treating the SAN as a single point of failure without redundant paths and power, and skipping failure testing. Validating the full chain from cameras and network to servers and storage helps avoid these issues.

Can I add capacity to a CCTV SAN later without disrupting recording?

Many enterprise SAN platforms support adding drive enclosures or expanding pools online, but the details vary by vendor, RAID configuration and VMS. Check the expansion model during the design phase, including whether new capacity can be added to existing volumes or only as new volumes. Planning growth headroom up front reduces the need for disruptive changes later.

How can I test my design before buying the hardware?

Use a simulation or benchmarking tool to model your camera network, including resolutions, codecs, frame rates and analytics workloads, and measure throughput limits on candidate servers and storage. Tools such as Secure Logiq’s Logiqal Benchmark are designed for this purpose. Combine simulation with a proof-of-concept that includes failure scenarios such as pulling a drive or failing a path.

Related Articles