GateRev

Engineering evidence

Every measurement behind GateRev, each with the conditions it was taken under — load, host, data size and date. A benchmark without its conditions is a slogan.

Numbers, with the conditions attached

A benchmark without its conditions is a slogan. Ours come with both.

p95 170 ms

Baseline: 200 registrations with 50 live boards (SSE) connected.

p95 344 ms

Re-measured under the same load — 200 registrations, 50 live boards — on a shared host at 117% saturation (loadavg 37.3 across 32 cores).

3798 tests

Measured 11 Oct 2026: a green gate, plus 286 end-to-end tests in a real browser and a smoke test of the production build that must answer before every deploy. The count is dated on purpose — it grows with every sprint, and an undated number would simply be wrong.

Every guard proven by a mutation

A test that stays green when the fix is removed is not evidence. Before a guard ships here, the thing it protects is deliberately broken and the guard must go red — then it is restored. On 28 Aug 2026 the guard requiring a state change and its audit record in one transaction was checked three ways: record moved outside the transaction, record written through the wrong client, and the fix reverted entirely. All three turned it red.

Restore drill: 1.61 s

Measured 24 Aug 2026: database restored from a nightly dump in 1.61 s (1.9 MB dump, 40 tables, 5,636 visits), vehicle photos in 0.013 s, with no data lost against the dump. Backups run nightly, so the recovery point is the last nightly backup, and every restore is cross-checked: the SHA-256 recorded in the restored database matches the file taken from the archive.

Built for failure

Designed for failure, per deployment. In GateRev Cloud, automatic authorisation runs over the connection between the site and the cloud; if that path goes down, the site switches to its manual gate procedure — the FAQ says the same. GateRev Onsite runs the application inside the customer's infrastructure, so an Internet outage alone does not stop it as long as the local server, network, ANPR integration and barrier controller are available — which does not mean every external integration stays available. In every deployment the barrier's local safety functions are independent of GateRev. GateRev Cloud decides centrally, with no edge agent to install or maintain at the gate; GateRev Onsite is the full GateRev application on your own servers. ANPR in GateRev is a consumer of your camera system: it reads plate events pushed by the customer's supported cameras and confirms them over N consecutive reads. The camera stays your system — it is not part of what we deliver — and every integration is commissioned on your device, at your gate distance and lighting.

Retention periods are read from the code that enforces them

The periods on our privacy page come from the same registry the deletion job reads, so the sentence shown to a driver always matches the schedule that deletes their data.

Both runs use the same load — 200 registrations with 50 live boards. The second was taken on a shared host running at 117% saturation, and GateRev still answered within 344 ms at p95. The measured code path has not changed between the two runs.

Last updated: 2026-10-12