Lead Forensics

Can multiple VMS recording servers use the same SAN storage infrastructure?

Yes, multiple VMS recording servers can use the same SAN storage infrastructure, provided each server writes to its own dedicated volume and the SAN has enough sustained write throughput for the combined load. Problems arise from shared contention, not from sharing itself. This article covers how to prevent conflicts, size shared SAN storage, choose a protocol and decide when a SAN is the wrong fit.

How do recording servers share a SAN without conflicts?

Recording servers avoid conflicts on a SAN by writing to separate, dedicated volumes rather than to the same file system. A block-level SAN presents storage as raw disks, so two servers writing to one standard NTFS volume without cluster-aware software will corrupt the data.

The key principle is exclusive ownership. Each recording server mounts its own LUN (logical unit number) and treats it as a local disk. The SAN controller handles the underlying physical drives, and LUN masking ensures that only the intended server can see each volume. This is why a SAN supports many recorders in practice: the sharing happens at the level of physical capacity and controllers, not at the level of files.

Shared-volume designs do exist, but they require a cluster file system or a VMS that explicitly supports shared archive access. Unless your VMS vendor documents this, assume one volume belongs to one recorder.

Should each VMS server get its own LUN?

Yes, in most surveillance designs each recording server should have its own LUN, or its own set of LUNs. Dedicated volumes prevent file system conflicts, make capacity easy to track per server and limit the damage if one volume develops a fault.

Several practical points shape the decision:

  • Isolation: A problem on one recorder’s volume, such as corruption or a full disk, does not spread to other recorders.
  • Capacity management: Each server’s retention can be planned and monitored independently.
  • Volume size: Very large single volumes can be slow to check or repair, so splitting a recorder’s archive across several moderately sized LUNs is often sensible.
  • Naming and documentation: A clear map of server to LUN to RAID group saves time during fault finding.

Check your VMS vendor’s guidance on maximum volume sizes and supported storage types before finalising the layout, as requirements vary between platforms.

What limits performance on shared SAN storage?

Performance on shared SAN storage is limited by sustained write throughput, controller capacity, the number of spindles behind each RAID group and the network or fabric between servers and storage. Video recording is a continuous sequential write workload, so peak specifications matter less than what the array can sustain around the clock.

The main constraints to evaluate are:

  • Combined bitrate: Add up the recording streams from every connected server, then allow headroom for playback, exports and analytics searches.
  • Controller and cache limits: A SAN controller shared by many recorders can become the bottleneck before the drives do.
  • RAID overhead and rebuilds: Parity RAID levels carry write overhead, and a rebuild after a drive failure reduces available throughput while it runs. Test the array under rebuild conditions, not only when healthy.
  • Mixed workloads: General IT arrays tuned for random database access behave differently from storage built for sequential video writes. Shared arrays also suffer when other applications compete for the same disks.
  • Analytics and metadata: AI analytics can add significant metadata writes and heavy read activity, which changes the load profile.

Do not rely on drive specifications alone. Performance depends on the whole chain, from the recording server and network to the controllers and disk groups. A design review that models the full camera load, such as Secure Logiq’s system design service, helps expose these limits before purchase rather than after deployment.

iSCSI or Fibre Channel for surveillance SAN?

Both iSCSI and Fibre Channel can support surveillance storage, and neither is universally better. Fibre Channel offers a dedicated, low-latency storage fabric but needs specialist switches and host bus adapters. iSCSI runs over standard Ethernet, which lowers cost and simplifies skills requirements, but demands careful network design.

When iSCSI makes sense

iSCSI suits projects where the team already manages Ethernet networks and wants to reuse that expertise. Good practice is to isolate storage traffic on its own VLAN or physical network, use sufficient link speeds and configure redundant paths. Poorly planned iSCSI that shares a congested network with camera traffic is a common cause of dropped recordings.

When Fibre Channel makes sense

Fibre Channel is more common in organisations that already operate a Fibre Channel fabric for other enterprise workloads, or where a dedicated storage network is a strict policy requirement. The trade-off is higher hardware cost and more specialist administration.

Whichever protocol you choose, confirm compatibility with your recording servers, operating system and VMS. Redundant paths, correctly configured multipathing and properly sized links matter more than the protocol label.

Does SAN sharing work with VMS failover?

SAN sharing can support VMS failover, but only when the design and the VMS both support it. In a typical failover setup, a standby recording server takes over a failed server’s cameras and needs access to that server’s archive volume or a defined replacement location.

Several points need verifying with your VMS vendor:

  • How failover accesses recordings: Some platforms continue recording to a new location and join the archives later, while others expect the standby to mount the original volume.
  • Volume ownership during failover: A volume must never be mounted read-write by two servers at once. Fencing or clean release of the LUN is essential.
  • Single point of failure: A SAN protects against recorder failure only if the SAN itself is resilient. Dual controllers, redundant fabrics or networks and redundant power are needed, and even then the array remains a shared dependency.

Remember that redundancy does not guarantee uninterrupted recording, and RAID is not a backup. Test failover and failback procedures during commissioning, as a configuration that looks sound on paper can fail in practice.

When is a SAN the wrong storage choice?

A SAN is the wrong choice when the added complexity, cost and shared dependency outweigh the benefits. For small and mid-sized deployments, or sites with predictable growth, recording servers with their own internal or directly attached storage are often simpler, faster to deploy and easier to support.

Consider alternatives in these situations:

  • Few recorders: The overhead of fabrics, switches and a SAN array rarely pays off for small server counts.
  • Limited storage skills on site: A SAN needs ongoing management, firmware maintenance and monitoring.
  • Remote or multi-site estates: Distributed recording close to the cameras can reduce network dependency.
  • Avoiding a shared failure domain: If a single storage fault could affect every recorder, independent storage per server may be safer.

A SAN earns its place where capacity must be pooled, expanded independently of compute, or integrated with an existing enterprise storage platform. Secure Logiq’s Qatar University SAN storage case study shows one example of a large deployment where this approach was applied. For projects with fewer recorders, purpose-built HD surveillance servers with integrated high-capacity storage are frequently the more practical route.

If you are weighing SAN against server-based storage for a specific project, contact the Secure Logiq team to discuss your camera count, retention requirements and resilience targets.

Frequently Asked Questions

How do I calculate how much SAN storage my recording servers need?

Start by totalling the bitrate of every camera across all recording servers, then multiply by your required retention period in seconds and convert to terabytes. Add overhead for RAID parity or hot spares, metadata from analytics, and 15-20% free space so volumes never fill completely. Many VMS vendors provide storage calculators, and it is wise to validate the result against your actual codec, frame rate and motion-recording settings rather than nominal values.

How do I get started with provisioning a SAN for multiple recording servers?

Begin with a design that maps each recording server to its own LUN or set of LUNs, backed by RAID groups sized for the combined sustained write load. Configure LUN masking or zoning so each server only sees its own volumes, set up redundant paths with multipathing, and document the server-to-LUN-to-RAID-group map. Before going live, run a sustained write test with realistic camera load, including a simulated drive rebuild.

What should I monitor after deploying shared SAN storage?

Track controller CPU and cache utilisation, per-LUN write latency, RAID group health, free capacity per volume and link or fabric errors. Set alerts for degraded arrays and rebuilds, since throughput drops during those events and can cause dropped recordings. Also monitor the VMS itself for recording gaps or archive warnings, as these often reveal storage problems before the SAN reports a fault.

What should I do if recordings start dropping after adding more servers to the SAN?

First check whether the combined write load now exceeds what the controllers or RAID groups can sustain, and look for latency spikes on the affected LUNs. Then review the network or fabric for congestion, particularly if iSCSI traffic shares links with camera streams, and confirm no rebuild or other workload is competing for the same disks. If the array is at its limit, options include moving recorders to separate RAID groups, adding controllers or spindles, or reducing bitrates.

Can I mix surveillance and general IT workloads on the same SAN?

It is possible, but it carries risk because surveillance is a constant sequential write load while general IT workloads are often random and bursty. To reduce contention, dedicate separate RAID groups or disk pools to video, apply quality-of-service limits if the array supports them, and test under peak load from both workloads. If you cannot guarantee isolation, a dedicated surveillance storage platform is usually safer.

How often should I test VMS failover on SAN storage?

Test failover and failback during commissioning, after any major change to the VMS, SAN firmware, network or LUN layout, and on a regular schedule such as annually or more often for critical sites. Each test should confirm that cameras resume recording, that the volume is never mounted read-write by two servers, and that archives are accessible for playback afterwards. Record the results so you can compare behaviour over time.

What are the most common mistakes when sharing a SAN between recording servers?

Frequent mistakes include presenting one volume to several servers without a cluster file system, sizing for average rather than sustained and rebuild-time throughput, and running iSCSI over a congested shared network. Others are creating very large single volumes that are slow to repair, neglecting redundant controllers or paths, and treating RAID as a backup. A design review and thorough pre-deployment testing catch most of these issues early.

Related Articles