Deployed where the retention rules are strictest

These are the sectors where storage planning has already failed once, where the physical space has run out, and where the footage has to remain usable as evidence long after it was recorded.

Sectors

Eight environments we build for

The compression core is the same everywhere. What changes is the form factor, the power envelope and the compliance requirement.

Public safety & Safe City

City-scale video aggregation with retention that stands up as statutory evidence, and compression that survives audit review.

Telecom operators

Long-term subscriber video archiving at national scale, where storage growth is a direct hit to operating margin.

Rail & transit

Core transport hubs running 24/7, where a single point of failure is not an acceptable outcome and footage is often subject to legal hold.

Smart campus & parks

Dozens of distributed buildings and gates under one operating picture, usually with no central machine room.

Finance & retail

Branch, ATM and point-of-sale networks with long audit trails and rigid dedicated-line budgets.

Healthcare

Ward, corridor and theatre coverage where footage is frequently needed to resolve disputes months after the event.

Industrial IoT

Production lines under strict safety supervision, needing both continuous storage and live compliance alerting.

Energy & utilities

Unmanned pumping stations, substations and remote poles on solar power and constrained wireless bandwidth.

Case studies

What changed on site

Every project below replaced a conventional server-and-array or IPC-and-NVR architecture. In each case the incumbent video platform stayed in place.

Industrial manufacturing

80 streams, one weak-current cabinet, no machine room

Constraints

  • Infrastructure limit. 80 channels of 1080P needed to be onboarded, but the site had only a weak-current cabinet — no standard equipment room.
  • Safety requirements. No open flame permitted in the workshop, so long-term recording plus AI alerting for helmets, smoking and absence from post was mandatory.
  • Budget ceiling. A conventional server plus storage array build was unaffordable.

What we deployed

  • A single small appliance with a built-in VMS, connected directly at the cabinet.
  • Six compute cards driving fire and smoke warning, helmet detection and smoking detection.
  • Compression at H.264 90%, allowing the full 80-stream estate to be stored on four hard drives.

Outcome

  • Drive bays reduced from 16 to 2.
  • Equipment removed from the machine room entirely — now housed in the weak-current cabinet.
  • Live AI alerting finally put the safety monitoring wall to work.
70% less space and 60% less energy than the equivalent conventional build, with no separate AI server and no large storage array.
“The room used to be full of equipment. Now all 80 channels plus AI alerting sit in the weak-current cabinet. We freed the space and the cost, and the safety wall is finally doing its job.”Security Manager, precision machining plant
Justice & corrections

Nearly 1,000 streams converged, retention extended six-fold

Constraints

  • Extreme concurrency. Almost a thousand round-the-clock streams arriving at once. The conventional architecture required so many nodes that network and compute pressure became the binding constraint.
  • Storage cost. High-definition capture consumed capacity so fast that 30-day retention was already the physical and budgetary limit.
  • Zero tolerance for downtime. The facility demands uninterrupted, gap-free coverage, so single-node failure risk was unacceptable.

What we deployed

  • SWIFTServer S Series standard and enhanced nodes, each carrying up to 384 concurrent 1080P streams.
  • Multi-node cluster hot standby with service-level transparent failover.
  • Ten-to-one compression on the landed archive with full preservation of analytical features.

Outcome

  • Physical carrier pressure removed by consolidating streams per node.
  • Retention extended from 30 days to six months.
  • Downstream AI structured analysis unaffected by compression.
Compression reduced footage to roughly one tenth of its original footprint while preserving picture quality and analytical usability, delivering a 70% reduction in supporting build cost.
Banking

Eight branches, 275 streams, 120-day retention without new NVRs

Constraints

  • 275 streams across eight branch sites, previously served by stacked conventional NVRs.
  • A 120-day retention requirement that local storage could not physically hold.
  • Frequent retrieval over the bank's dedicated line, with persistent stuttering and timeouts.

What we deployed

  • SWIFTBox P8/P16 compression gateways installed in each branch equipment room.
  • Edge compression of roughly ten to one, returned over the dedicated line using national-standard cascading.
  • No change to the existing monitoring architecture or the head-office platform.

Outcome

  • High-cost local NVR deployment eliminated entirely.
  • Head-office retrieval of multiple streams now opens instantly.
  • Centralised unified management across all eight sites.
Overall hardware cost fell substantially, with electricity and manual maintenance costs dropping by more than 100,000 RMB on the comparison project.
Pharmacy retail

100+ stores brought into 90-day medical-insurance compliance

Constraints

  • More than 100 pharmacy branches, each with two to four cameras.
  • A mandatory 90-day archiving rule imposed by the medical insurance authority.
  • Hardware procurement cost rising fast, difficult maintenance, and frequent errors on cross-region retrieval from head office.

What we deployed

  • SWIFTBox P2/P4 edge appliances replacing conventional IPC-plus-NVR installations.
  • Streams compressed at the edge, then forwarded directly to head office and central storage.
  • Built-in micro-storage for local upload buffering, with dual cloud-edge stream policy.

Outcome

  • Compliance audit passed in full.
  • Construction and hardware procurement cost reduced significantly.
  • Narrowband public-network retrieval now handles multiple concurrent streams without stuttering.
The NVR layer was removed entirely — fewer devices, lower cost and a simpler estate to maintain, while every store met the 90-day retention mandate.
Remote & off-grid

Single-source sites on solar power and 4G backhaul

Constraints

  • Remote water monitoring stations and field poles with no mains power.
  • Harsh outdoor temperature ranges and weather exposure.
  • 4G/5G data tariffs that made continuous upload economically unviable.

What we deployed

  • SWIFTBox P1 single-stream gateways, drawing 3.6 W and weighing 150 grams.
  • Compression of roughly ten to one, taking a 1080P stream to about 0.5 Mbps.
  • Packet-loss tolerance designed to avoid screen corruption on unstable links.

Outcome

  • Bandwidth tariff reduced by roughly 90% per site.
  • Solar sizing unchanged despite continuous operation.
  • Timestamp alignment preserved so that footage remains usable as evidence.
Up to 10:1 compression at 3.6 W, making continuously monitored off-grid sites financially practical for the first time.
Telecom & carrier archive

National video archive without expanding the data centre

Constraints

  • Continually expanding archive of subscriber and infrastructure video.
  • Central storage footprint growing every quarter without a matching revenue line.
  • Strict internal security requirements precluding external network access.

What we deployed

  • X-Stream Archive running on general-purpose CPU hosts already in the estate.
  • Delayed scanning with automatic scene classification and template matching.
  • In-place replacement with full four-invariant verification on every file.

Outcome

  • No additional hardware purchased on the archive path.
  • Archive footprint reduced to approximately one tenth.
  • Fully isolated internal deployment with no outbound connectivity.
Roughly 2.5 PB of storage released on the reference deployment, achieved entirely in software on existing compute.
Repeatable pattern

What these projects have in common

Retention requirement was the trigger

In every case the project started because a retention period was extended beyond what the existing storage could physically hold.

The incumbent platform stayed

No client replaced their video management platform. Our layer sits underneath it and remains invisible to its workflows.

The comparison was against expansion

Every business case was measured against buying more arrays, more racks and more power — not against doing nothing.

Verification decided the deal

Independent inspection reports confirming unchanged resolution, frame count, duration and codec were what made approval possible.

Which of these looks like your site?

Send us the closest match plus your camera count and retention target. We will come back with a sizing estimate and a deployment sketch.

Request a consultationSee the architecture