Two cameras with the same resolution and frame rate can produce very different traffic. Motion, foliage, rain, image noise, wide dynamic range, PTZ movement and the encoder's rate-control method all change the number of bits required. An average bitrate may estimate long-term storage, but it does not prove that a switch uplink, recorder interface or disk pipeline can absorb short bursts from many cameras at once.
This worksheet turns bitrate from a tender assumption into a measured commissioning record for campuses, hotels, hospitals, warehouses, schools and commercial buildings across Telangana, Andhra Pradesh and South India.
1. Freeze the production stream inventory
Measure the streams that will actually run after handover. A camera can publish a main recording stream, low-resolution live stream, mobile profile, analytics stream and direct integration stream at the same time. The VMS may request settings that differ from the camera web page.
| Record for every profile | Why it matters |
|---|---|
| Camera ID, model, firmware and VMS driver | Makes the result repeatable after upgrades or replacement |
| Codec, resolution and frames per second | Defines the basic encoder workload and delivered detail |
| Rate control, target and maximum bitrate | Separates an average storage target from an immediate ceiling |
| GOP or GOV length and dynamic GOP state | Controls keyframe frequency, seek behavior and recovery after loss |
| Smart codec, dynamic frame rate, WDR and noise reduction | These functions can alter traffic as the scene changes |
| Audio, metadata, consumers and schedule | Identifies added traffic, duplicate pulls and recording states |
2. Distinguish storage control from peak protection
Axis describes variable bitrate as preserving the selected compression while allowing traffic to vary with activity. Its average bitrate method manages a longer-term storage budget, while maximum bitrate protects the immediate network or recorder load. These are different controls. Averages can still include temporary overshoot, and an unrealistically low maximum can damage important motion detail.
Collect average bitrate over the full window, 95th-percentile bitrate from fixed sampling intervals, the highest observed interval, total VMS bytes, dropped packets, incomplete frames and recorder warnings.
95th-percentile bitrate ÷ average bitrate
Observed burst factor
Highest interval bitrate ÷ average bitrate
Record the sampling interval. A one-second peak and a five-minute average are not interchangeable.
3. Understand GOP before changing it
A GOP starts with an independent I-frame, followed by frames that describe changes. More frequent I-frames improve random access and limit how long corruption can persist after packet loss, but increase bitrate. A longer GOP can reduce traffic in a quiet view, yet seeking and recovery depend on the next suitable keyframe.
Frames from one I-frame to the next ÷ configured frames per second
Example: a 50-frame interval at 25 fps gives a 2-second GOP. Confirm how the manufacturer defines the setting, because some interfaces count only predicted frames.
Seek to arbitrary times, reverse and frame-step where supported, start live view mid-GOP, and create a short controlled packet-loss event during an approved test window. Record time to a clean image. With dynamic GOP, repeat while the scene is still and moving.
4. Build a real scene challenge set
Measure each critical camera in likely conditions. Hyderabad and coastal Andhra installations can see strong daylight contrast, monsoon rain, reflective wet roads, insects near infrared illumination and low-light noise. Do not simulate every condition with an operator waving at the lens.
| Scene state | Observe | Evidence |
|---|---|---|
| Quiet baseline | Noise, dynamic GOP and frame-rate behavior | Average and keyframe interval |
| Normal traffic | Expected people, vehicles, lifts or gates | Average, 95th percentile and useful detail |
| Dense motion | Crowd, vehicle queue or shift change | Peak, recorder alarms and playback |
| Fine random motion | Rain, leaves, flags, water or escalators | Burst factor and artifacts |
| Low light and IR | Gain noise, insects, headlights and mode transition | Night peak and identification task |
| PTZ and WDR | Movement, zoom and backlit entrances | Peak traffic and image recovery |
5. Calculate storage from measured traffic
Use the measured full-day average for a continuously recorded stream, or separately account for each schedule state. Do not multiply a quiet daytime sample across an entire retention period.
Average bitrate in Mbps × 86,400 ÷ 8 ÷ 1,000
Required video capacity in TB
Sum of daily camera volumes × retention days ÷ 1,000
Add project-approved allowances for filesystem, database, audio, metadata, RAID behavior, growth and unavailable capacity. Do not count parity or hot spares as usable video capacity.
For event recording, calculate background or keyframe-only recording, full recording during events, and expected event duration separately. Validate the estimate against actual VMS bytes. Milestone notes that lowering frame rate may save less storage than expected because I-frames can form a large share of a quiet GOP.
6. Size shared paths for concurrency
Cameras do not peak independently when the cause is shared. Rain, a shift change, power-restoration restart or an evacuation can raise many streams together. Aggregate measurements by switch, uplink, recording server and storage group.
Sum of accepted 95th-percentile bitrates in the shared scenario
+ live-view and secondary-stream demand
+ edge retrieval or export traffic
+ protocol and project headroom
Recorder ingest utilization
Concurrent design load ÷ verified recorder ingest capacity × 100
Keep headroom explicit and owner-approved. Check both directions on links carrying edge backfill, exports or viewing. Inspect interface drops, recorder NIC counters, queues, disk latency and VMS alarms during the same timestamped test.
7. Prove quality at the accepted ceiling
Compare footage with the operational task: detect intrusion, recognize a known person, read a plate where designed, inspect a cash desk or follow motion through a corridor. Review moving subjects frame by frame, because a clean paused background can hide compression loss on faces and plates.
If the maximum is reached, determine what the encoder sacrifices. Do not raise compression merely to flatten the graph. Recompose moving foliage, improve night lighting, tune exposure and noise reduction, or split an overloaded scene when that better preserves evidence.
8. Check viewers and integrations
Open intended control-room layouts, mobile clients, analytics and third-party integrations while recording continues. Confirm whether they reuse a VMS stream or create direct camera sessions. Watch total camera egress, decoder load and recorder ingest as users change layouts, scrub playback and export clips.
Repeat a representative test after camera reboot, recording-server restart and firmware or VMS upgrade. Average-rate algorithms may need time to learn the scene, so document the manufacturer establishment period and avoid signing off immediately after reset.
9. Commissioning worksheet
| Test | Record | Decision |
|---|---|---|
| Profile baseline | Codec, resolution, fps, rate control, target, maximum, GOP and smart codec | Matches design |
| Quiet and normal scenes | Average, 95th percentile, peak, I-frame interval and quality | Pass, deviation or retest |
| Worst credible scene | Peak, burst factor, dropped traffic, alarms and retained detail | Pass, deviation or retest |
| Aggregate group | Concurrent load, uplink use, recorder ingest and disk latency | Headroom accepted |
| Storage reconciliation | Estimated daily GB, actual VMS bytes, variance and explanation | Retention approved |
10. Decision checklist
- Every delivered profile and consumer is inventoried.
- Camera, firmware, VMS driver and recorder versions are recorded.
- Codec, resolution, frame rate and image settings match the design.
- Rate control, target and maximum bitrate are distinguished.
- GOP duration is calculated and verified from the delivered stream.
- Seek, playback and recovery behavior are tested.
- Quiet, normal, dense-motion, low-light and weather-sensitive scenes are sampled.
- Average, 95th-percentile and peak values use a documented interval.
- Peak-to-average and burst factors are calculated.
- Forensic detail is reviewed during motion.
- Smart codec and dynamic GOP compatibility are verified.
- Storage is calculated from representative averages.
- VMS bytes are reconciled with the estimate.
- Shared-cause concurrency is tested by switch and recorder group.
- Live view, export and edge retrieval are included.
- Packet drops, recorder queues and disk latency are checked.
- Configurations, samples and timestamped logs are handed over.
- Firmware, VMS, camera position and lighting changes trigger retesting.
11. Handover package
Provide the camera schedule, production stream matrix, approved scene challenges, measurement interval, per-camera statistics, aggregate group results, storage workbook, sample recordings, quality decisions, switch and recorder counter captures, deviations, final configuration exports, version inventory and regression triggers. State whether results came from camera telemetry, VMS statistics, network counters or packet analysis.
Primary technical references
- Axis: Bitrate control for IP video
- Axis VAPIX: Rate control
- Axis: Zipstream technology
- Milestone XProtect: Storage configuration and GOP behavior
- Milestone XProtect 2026 R1: Enable keyframe recording
- Hanwha Vision: WiseStream advanced video data reduction technology
Commissioning CCTV bandwidth and storage?
Share the camera schedule, VMS platform, retention target, topology and difficult scenes. Cyber FTTH Mart can help structure measurements, aggregate load checks and handover evidence.
Discuss CCTV commissioning ↗