Summary
- JANOG58 used a local-5G installation to serve sponsor booths in the annex of the Ehime Prefectural Culture Hall during the 15–17 July meeting in Matsuyama.
- The official report says this was JANOG’s first operation of an actual local-5G system using an experimental test-station radio licence, with Ehime CATV as licensee.
- The disclosed service path ran from a Celona 5G LAN base station and an omnidirectional antenna to SIM-equipped CPE routers placed in sponsor rooms.
- The report records roughly 30GB uplink and 140GB downlink traffic over three days, but publishes neither throughput, latency, concurrent-user count nor a comparison with Wi-Fi.
The annex at a network-operator meeting is usually a facilities problem. At JANOG58 it became a radio-operations problem.
The meeting had originally planned to use only the main building of the Ehime Prefectural Culture Hall in Matsuyama. Sponsor demand exceeded that plan, so the host, Arista Networks Japan, extended booths into the separate annex. That created a compressed requirement: give the annex a wireless service comparable in purpose to the sponsor area in the main hall, without treating a provisional room layout as an excuse to skip the operational work that licensed spectrum demands.
The 13 August JANOG58 newsletter describes the answer. Arista, Ehime CATV and Sojitz Tech Innovation built a local-5G environment in the annex and used it to provide wireless access to sponsor booths during 15–17 July. The report calls it JANOG’s first real operation of a local-5G system under an experimental test-station licence. That distinction matters. A conference demo can show a radio link; a live, licensed service has to connect spectrum authorisation, installation, subscriber provisioning and day-of-operation monitoring into one accountable chain.
A venue constraint became a control surface
The report identifies Ehime CATV as the licensee and the annex as the base-station area. It says the parties consulted the Shikoku Bureau of Telecommunications, prepared the required material and obtained the experimental-station licence. The source describes a roughly two-month process including prior consultation, but it does not publish the application file, exact dates, spectrum parameters or any interference-coordination record. The safe conclusion is therefore narrow: the service was operated under an authorised experimental arrangement, not that licensing is universally quick or simple.
That caution is useful for conference organisers. Local 5G is not merely a different access-point choice. Its operating surface includes the named licensee, the authorised area, radio-qualified personnel, the devices permitted to attach and the boundary between a venue’s temporary floor plan and a regulated radio installation. Those items decide whether a last-minute capacity extension remains a managed service instead of an unowned technical experiment.
The physical design was deliberately compact. JANOG58 used Celona 5G LAN equipment, described by the report as an integrated CU/DU/RU platform. An omnidirectional antenna was suspended to reach across the annex. The meeting did not ask every exhibitor to manage a radio modem directly: 5G-capable CPE routers with local-5G SIMs were placed in the relevant rooms, giving booths a conventional local-network handoff. A local-5G wearable camera also appeared as a demonstration device.
This is a more informative architecture than the phrase “private 5G at a conference” suggests. The organisers turned a cellular access layer into a managed backhaul edge for individual booth spaces. The operational challenge therefore moved to provisioning and CPE placement: which room receives a router, which SIM belongs in it, which support boundary applies when a booth loses access, and how the radio path is observed without confusing it with the downstream LAN.
The public record is a deployment record, not a performance benchmark
JANOG58 published a small but concrete traffic record. For 15 July, one CPE router showed 11.46GB uplink and 27.15GB downlink, or 38.6GB in total. Across the three-day meeting, the report gives approximately 30GB uplink and 140GB downlink—about 170GB altogether. It also says sponsors described the service as comfortable and stable, and that the mission completed without a major trouble report.
Those are meaningful operational observations. They show that the installation carried real conference traffic rather than remaining a static exhibit. They do not establish a general performance claim. The newsletter does not disclose air-interface bandwidth, channel plan, signal measurements, latency, packet loss, peak rate, the number of active CPE routers, connected users, traffic mix or a matched Wi-Fi baseline. Nor does it say how much of the traffic was attributable to exhibitors rather than demonstration devices or operational staff.
The distinction is important because local 5G procurement can too easily collapse into a slogan about speed. JANOG58’s useful lesson is instead about integration. A licensed radio service can be made legible to temporary tenants through CPE, while keeping the radio-side responsibilities explicit. The 170GB total is evidence of use; it is not evidence that local 5G is automatically superior to every venue Wi-Fi design.
What the next deployment should disclose
The next conference-scale trial could make the operating case considerably stronger without becoming a marketing exercise. It should publish the licensed spectrum block and authorised geography where disclosure is appropriate; the number and placement logic of base stations and CPE units; concurrent attachment counts; radio and application availability; median and tail latency; loss; handover or reconnection behaviour; incident tickets; and a comparable service baseline. It should also separate exhibitor demand from test traffic.
Those measurements would allow operators to judge two different questions that the JANOG58 report rightly keeps separate. First: did the service operate safely and reliably enough for the annex? On the public account, it did. Second: what conditions would justify repeating it instead of extending wired or Wi-Fi infrastructure? The available evidence does not yet answer that.
JANOG58 nevertheless adds a valuable operational fact to the NOG record. When venue capacity forced an unplanned wireless expansion, the response was not a generic connectivity promise. It was a real, licensed and monitored local-5G deployment with a defined service handoff. For a community that often debates network design in abstractions, that is the part worth retaining: the difficult work was not simply emitting a signal, but making radio authorisation, equipment, tenant access and operations meet in one temporary building.