How does a write bottleneck affect recorded footage?
A write bottleneck causes dropped frames, gaps in recordings and corrupted or incomplete video files. When the storage cannot absorb data as fast as cameras send it, the recording server queues the surplus in memory. Once that queue is full, the VMS discards frames, and the footage you needed may simply not exist.
The damage is rarely uniform. Because bitrate rises with scene complexity, a system may record acceptably overnight and fail during a busy period such as an incident, a crowded event or a shift change. Playback may look normal at a glance while frame rates in the archive are far lower than configured.
Other consequences include recording server instability, VMS database errors and failed or delayed exports. Analytics workloads can suffer too, because metadata writes compete with video writes for the same storage resources. A write bottleneck therefore undermines evidential value, which is the very purpose of surveillance storage.
What causes write speeds to fall behind video streams?
Write speeds fall behind when the aggregate camera bitrate exceeds the sustained write throughput of the recording path. Typical causes include undersized disk arrays, RAID overhead, RAID rebuilds, network congestion, shared storage contention and using drives or controllers not designed for continuous sequential video writes.
The most common culprits fall into a few groups:
- Underestimated bitrates: Designs based on average rather than peak bitrates, or on camera datasheet figures that ignore high motion, low-light noise or higher frame rates added after commissioning.
- Drive and array limitations: Too few spindles for the required throughput, desktop-class drives not built for 24/7 surveillance duty, or a RAID level whose parity calculations reduce effective write performance.
- Degraded or rebuilding arrays: A failed drive or an active rebuild consumes controller and disk resources, often reducing available write headroom just when the system is already under stress.
- Mixed workloads: Video, metadata, operating system activity, exports and playback sharing the same volumes can fragment disk access and reduce sequential efficiency.
- Network and shared storage limits: Saturated switch uplinks, undersized network interfaces or shared SAN and NAS paths where multiple recording servers compete for the same resources.
It is worth stressing that capacity and throughput are different design questions. An array can hold weeks of footage and still be unable to ingest the live stream rate.
What are the warning signs of insufficient write performance?
The clearest warning signs are VMS alerts about dropped frames or recording failures, rising disk queue lengths, high disk latency and recorded frame rates lower than configured. Intermittent gaps in archived footage, particularly during busy periods, strongly suggest the storage cannot sustain the incoming load.
Check these indicators on the recording server:
- VMS event logs showing buffer overflows, database write errors or skipped frames
- Sustained high disk queue length or write latency in the operating system’s performance monitor
- Controller or array alerts for degraded volumes, rebuilds or cache problems
- Memory use on the recording server climbing steadily as buffers fill
- Footage that plays back with irregular frame timing or missing segments
- Problems that worsen when analytics, exports or multiple client viewers are active
Do not rely on a single quiet-hour test. Review behaviour at peak activity, during exports and while an array is degraded, since these are the conditions that expose marginal designs.
How do you calculate the write throughput a CCTV system needs?
Add up the bitrates of every camera recording to the server or storage volume, then include headroom for peaks, metadata and overhead. Total write throughput is the sum of all stream bitrates, converted into the same units as the storage specification, with a safety margin applied.
The method is straightforward:
- Establish the bitrate per camera using the actual resolution, codec, frame rate and expected scene complexity, ideally measured on site rather than taken from datasheets.
- Sum the streams across all cameras per recording server and per storage volume.
- Convert units carefully. Camera bitrates are normally quoted in megabits per second, while disk throughput is usually quoted in megabytes per second. Divide megabits by eight to get megabytes.
- Add allowances for peak bitrate variation, secondary streams if recorded, audio, analytics metadata and any redundant or mirrored recording.
- Apply headroom so the array still performs adequately when degraded or rebuilding.
As a hypothetical illustration only: 200 cameras each averaging 8 Mbps produce 1,600 Mbps, or roughly 200 MB per second of continuous writes before any allowance for peaks, metadata or overhead. Real requirements will differ with codec, scene and configuration.
Manual calculation can miss interactions between camera settings, server limits and storage behaviour. Secure Logiq’s system design service produces bandwidth and storage calculation reports, and the Logiqal Benchmark utility is designed to simulate multi-camera workloads so that throughput limits can be validated before physical deployment.
How can write speed bottlenecks be prevented or fixed?
Prevent bottlenecks by designing for peak sustained throughput with headroom, using surveillance-grade drives and hardware RAID, separating the operating system from video storage and validating the design before deployment. Fix existing problems by locating the constrained component, then reducing load, adding spindles or redistributing cameras.
Preventing bottlenecks at the design stage
Size for peak bitrates, not averages, and keep headroom for degraded array operation. Use enterprise drives rated for continuous surveillance writes, and choose a RAID level that balances protection and write performance for the load. Secure Logiq enterprise servers, for example, isolate the operating system on dedicated mirrored SSDs so that system activity does not compete with video recording, and use battery-backed, hardware-accelerated RAID with hot-swappable spares. Understanding the server selection criteria early helps avoid undersized platforms.
Fixing a system that is already dropping frames
Start by finding the actual constraint. Check array health, controller cache status, network utilisation and per-volume latency. Then consider practical remedies:
- Lower camera bitrates or frame rates where the operational requirement allows
- Spread cameras across additional recording servers or volumes
- Add drives or expand the array to increase aggregate throughput
- Replace unsuitable drives or controllers with surveillance-grade components
- Remove network bottlenecks, such as saturated uplinks or single interfaces
- Move exports and heavy playback away from the primary recording volumes
Large or unusual deployments may need a purpose-built approach. Secure Logiq’s bespoke solutions and range of HD surveillance servers are aimed at demanding recording workloads of this kind.
If you suspect your surveillance storage cannot sustain its live camera load, or you are planning a new deployment, speak to Secure Logiq about your throughput requirements before specifying hardware.
Frequently Asked Questions
How do I start testing whether my existing CCTV storage can handle its current camera load?
Begin by recording the actual per-camera bitrates from your VMS during both quiet and busy periods, then sum them and compare the total against the sustained sequential write rating of your array. At the same time, monitor disk queue length, write latency and recording server memory in your operating system’s performance monitor during peak activity. If latency or queue length stays high, or buffers keep growing, the storage is likely the bottleneck.
Can a benchmark tool like CrystalDiskMark or Iometer tell me if my array is good enough for CCTV?
Generic benchmarks give a useful baseline for raw disk speed, but they rarely reflect a real surveillance workload of many concurrent sequential streams plus metadata, exports and playback. For a more realistic result, use a tool that can simulate multi-camera workloads, such as the Logiqal Benchmark utility mentioned above, and test under degraded and rebuilding conditions as well as with a healthy array.
How much headroom should I build into my write throughput calculation?
There is no single figure, but a common approach is to size for peak rather than average bitrates and then add a safety margin so the array still copes when degraded or rebuilding. Allow extra for secondary streams, audio, analytics metadata and future frame rate or camera additions. The more unpredictable the scenes, the larger the margin should be.
What should I do if frames are being dropped while my array is rebuilding after a drive failure?
Check the controller’s rebuild priority setting first, as lowering it can free write capacity for live recording, though it will lengthen the rebuild window. In the short term, consider temporarily reducing frame rates or bitrates on non-critical cameras and pausing exports or heavy playback. Longer term, design with enough headroom and hot-swappable spares so a rebuild does not push the system past its limits.
Will switching to H.265 or lowering resolution solve a write bottleneck?
It can reduce the load significantly, since lower bitrates mean less data to write, but it is only one lever and may compromise image quality or evidential value. H.265 and smart codec options also depend on camera and VMS support, and bitrates still spike in complex scenes. Treat bitrate reduction as a complement to proper storage sizing, not a replacement for it.
Is adding more drives always the right fix for dropped frames?
Not always. More spindles increase aggregate throughput, but if the real constraint is a saturated network uplink, a single network interface, an overloaded RAID controller or competing workloads on the same volume, extra drives will not help. Identify the constrained component first, then decide between adding drives, redistributing cameras across servers or volumes, or upgrading controllers and networking.
What are the most common mistakes when designing storage for IP camera systems?
The most frequent mistakes are sizing for capacity but not throughput, relying on datasheet or average bitrates, using desktop-class drives, and putting the operating system, video, exports and playback on the same volumes. Another is validating the design only when the array is healthy and activity is low. Testing at peak load and in degraded states before deployment avoids most of these problems.