Separating video recording servers from centralised storage lets each layer do one job: recording servers ingest and manage camera streams, while a shared storage platform holds the video. This removes single-box limits on capacity, throughput and resilience, and lets you scale storage without buying more compute. This article explains where combined servers struggle, how a split design performs, and when it is worth the extra complexity.
What bottlenecks does a combined recording and storage server create?
A combined server forces camera ingest, VMS processing, analytics, client streaming and disk writes to compete for the same CPU, memory, network ports and drive bays. The first resource to saturate caps the whole system, and the only fix is usually another complete server.
In practice, these are the pressure points designers meet most often:
- Chassis capacity: the number of drive bays fixes the maximum retention, so extending retention can mean adding another recording server even when compute is idle.
- Stranded resources: one server may run out of storage while another has spare capacity, and the two cannot share it.
- Mixed workloads: sustained sequential writes, playback reads and analytics processing all compete for the same disks and controllers.
- Network interface limits: incoming camera traffic, live client views and exports share the same server connections.
- Server sprawl: adding servers purely to gain disk space increases rack space, power draw and cooling demand.
Purpose-built surveillance servers reduce these issues through hardware RAID, dedicated operating system drives and optimised configurations. Even so, a single chassis remains a finite pool of compute, ports and bays. For smaller and mid-sized systems that is often perfectly adequate, which is why the question of when to split matters, and we return to it below.
How does a split architecture improve resilience and uptime?
A split architecture improves resilience by removing the one-to-one link between a recording server and its disks. If a recording server fails, its storage remains intact and another server can take over; if a storage component fails, redundancy within the storage platform protects the video already recorded.
Resilience gains come from several design choices that become practical once storage is shared:
- Recording server failover: a standby or failover recording server can be pointed at the same storage, subject to what your VMS supports.
- Storage redundancy: enterprise storage typically combines RAID, hot spares and redundant power supplies, so a drive or power fault does not stop recording.
- Independent maintenance: recording servers can be patched, upgraded or replaced without moving large volumes of video.
- Fault isolation: a software fault on one recording server does not take the archive offline.
It is worth being precise about what this does and does not achieve. RAID protects against drive failure but is not a backup, and redundancy reduces the likelihood of interruption rather than guaranteeing uninterrupted recording. A shared storage platform can also become a point of concentration, so the storage itself needs redundant controllers, power and network paths. Failover behaviour depends on the VMS in use, so confirm what your chosen platform supports before designing around it.
How do you size storage independently of recording servers?
Size recording servers from camera count, bitrate, analytics load and client activity. Size storage from the aggregate recording bitrate, retention period, redundancy overhead and growth allowance. Treating the two calculations separately lets each scale on its own terms, which is the central practical benefit of a split design.
Calculating the storage requirement
Start with the total sustained bitrate of all cameras that will record, then multiply by the retention period to get raw video volume. As a simple hypothetical illustration, 500 cameras each recording at 4 Mbps produce a total of 2,000 Mbps, or 250 MB per second, before overheads. Over 30 days of continuous recording that equals roughly 650 TB of raw video. These figures are illustrative assumptions only, and real bitrates vary with codec, resolution, frame rate, scene activity and recording mode.
On top of raw volume, add RAID parity and hot spare overhead, filesystem overhead, and headroom for growth and for retention changes. Where AI analytics are in use, allow for metadata storage as well.
Sizing the recording servers
Recording server sizing follows the number of camera streams each server can ingest, the VMS licensing and limits, any analytics processing and the number of concurrent operators. Because storage is no longer tied to the chassis, you can choose server specifications on compute and network needs alone. Tools such as Secure Logiq’s Logiqal Benchmark simulate camera and analytics workloads to validate throughput before physical deployment, and the company’s server selection guidance covers the main sizing inputs.
What network and bandwidth demands does centralised storage add?
Centralised storage adds a second network path: video now travels from cameras to the recording server, then across a storage network to the storage platform. That storage link must sustain the combined write rate of every recording server plus playback, export and rebuild traffic, with headroom and redundant paths.
Key considerations include:
- Dedicated storage network: keeping storage traffic apart from camera and management networks avoids contention and makes performance easier to predict.
- Link capacity: the aggregate write rate must fit comfortably within the storage fabric, with allowance for reads during playback and for RAID rebuilds.
- Path redundancy: redundant switches, ports and cabling prevent a single link failure from interrupting recording.
- Latency and sustained writes: continuous sequential video writes behave differently from typical IT workloads, so storage must be assessed on sustained performance, not peak figures.
The additional switching, cabling and configuration are real costs. A common mistake is to size the storage array generously but under-provision the network between servers and storage, which then becomes the new bottleneck.
What’s the difference between centralised storage and local server storage?
Local server storage places disks inside, or directly attached to, each recording server, so capacity and performance belong to that server alone. Centralised storage, such as a SAN, pools disks into a shared platform that multiple recording servers use over a dedicated network, allowing capacity to be allocated flexibly and scaled independently.
The main trade-offs compare as follows:
- Scalability: local storage grows by adding servers or drive bays; centralised storage grows by adding capacity to the shared platform.
- Utilisation: local storage can leave capacity stranded on individual servers; centralised storage pools it.
- Complexity: local storage is simpler to design, deploy and troubleshoot; centralised storage adds a storage network and more components to manage.
- Failover: local storage ties recordings to one server; centralised storage makes recording server failover more practical.
- Cost profile: local storage is often lower in initial cost for small systems; centralised storage can reduce rack space, power and duplicated hardware at scale.
Neither approach is universally better. A SAN is not inherently the right choice for every CCTV deployment, and a well-specified, purpose-built local storage design remains the sensible answer for many projects. Secure Logiq’s Qatar University SAN storage case study shows how a centralised approach was applied in a large campus environment.
When does separating recording from storage make sense?
Separating recording from storage makes sense when a project needs storage to scale independently of compute, when recording server failover is a requirement, or when long retention across many cameras would otherwise cause server sprawl. For smaller, self-contained systems, combined servers are usually simpler and more economical.
A split design is worth considering when:
- Camera counts, bitrates or retention periods are large and likely to grow.
- Retention requirements are long, or change between sites or camera groups.
- Several recording servers need to share one pool of capacity.
- The VMS and project specification call for failover or high availability.
- Rack space, power and cooling are constrained and consolidation is a priority.
- AI analytics generate significant metadata and archive demands.
A combined server remains the better fit where the camera count is modest, retention is predictable, the network is simple and the support team prefers fewer moving parts. The deciding factor is rarely the technology alone. It is the balance between resilience requirements, growth expectations and the operational capability to manage a storage network. Secure Logiq offers HD surveillance servers for combined and distributed designs, and its newer Logiqal Core Architecture and Logiqal Video Volume, introduced in its Intersec 2026 announcement, address large-scale video storage requirements.
If you are weighing a split architecture for a new or expanding project, the Secure Logiq system design team can help you model recording and storage requirements before committing to hardware.
Frequently Asked Questions
How do I get started if I want to move from combined servers to a split recording and storage design?
Begin by auditing your current system: record the number of cameras, actual sustained bitrates, retention targets and growth plans per site. Then size recording servers (compute and network) and storage (aggregate bitrate × retention, plus RAID, spare and growth overhead) as two separate calculations. Before buying hardware, validate the design with a workload simulation tool such as Logiqal Benchmark, and confirm with your VMS vendor which failover and shared-storage configurations are supported.
Can I migrate gradually, or does it have to be a full replacement?
A phased migration is often possible. You can introduce centralised storage for new cameras or new sites first, while existing combined servers continue recording locally until they reach end of life. Check how your VMS handles mixed storage targets and how existing archives will be retained or moved, and plan the storage network upfront so it can support later phases without rework.
What happens to recording if the storage network link fails?
If there is only a single path between recording servers and storage, a link, switch or port failure will interrupt recording. This is why redundant switches, ports, cabling and multipath configurations are recommended. Also test failure scenarios during commissioning, and check whether your VMS can buffer or fall back to local disk temporarily, as this varies between platforms.
Is RAID on the shared storage platform enough to protect my footage?
No. RAID protects against drive failures but it is not a backup, and it does not protect against accidental deletion, corruption, or loss of the whole storage platform. For critical footage, combine RAID and hot spares with redundant controllers and power, and consider replication or a separate archive for evidence that must be preserved.
What are the most common mistakes when designing a split architecture?
The most frequent error is under-provisioning the network between recording servers and storage while sizing the array generously, which simply moves the bottleneck. Others include sizing from peak rather than sustained write performance, forgetting RAID rebuild and playback traffic in bandwidth calculations, and omitting headroom for growth or analytics metadata. Using real bitrate measurements rather than assumed values also avoids costly under- or over-sizing.
How do I know if my team can support a centralised storage environment?
Consider whether you have staff or a support partner comfortable with storage networking, monitoring, firmware management and failure recovery. A split design adds components such as storage switches, controllers and multipath settings that need ongoing attention. If that capability is limited, a well-specified combined server design or a managed support arrangement may be the more reliable choice.
How should I plan for retention periods that differ between sites or camera groups?
Calculate storage per camera group using its own bitrate, recording mode and retention target, then add the totals together with overhead and growth allowance. A shared storage pool makes it easier to reallocate capacity when requirements change, rather than leaving unused space stranded on individual servers. Review the figures periodically, as codec changes, higher resolutions and new analytics can shift the requirement.