Site gateway — Field connectivity (fos-edge)
The gateway (fos-edge) runs on a Linux industrial PC at the site. It speaks the field protocols, checks every write before executing it, and keeps the data safe when the network is not.
Implemented, not yet validated on site: the protocol drivers, hot standby, multi-path uplink, redundant device paths and outage backfill on this page are covered by automated tests but have not yet been validated one by one on real equipment at a customer site; real devices are checked in a lab first.
Many protocols, one gateway
- BACnet/IP and BACnet/SC, Modbus TCP and RTU, OPC UA, IEC 61850, DNP3, IEC 60870-5-104, Siemens S7, EtherNet/IP, SNMP, MQTT, KNX, M-Bus and ONVIF.
- A network scan discovers devices, matches make and model against the model library and drafts the point list; nothing takes effect until a person confirms it.
- Device credentials (SNMP, OPC UA, MQTT and others) are entered on the platform, encrypted and write-only, and sent encrypted only to the gateway that needs them.
Receiving pushed events
- SNMP traps and informs (v1, v2c, v3), syslog (UDP, TCP, TLS) and HTTPS webhooks.
- Off by default and switched on per gateway by a manager; only device addresses registered in the point list are accepted.
- The alarms go to the alarm center; acknowledging them on the platform never changes the device.
Redundancy
- Hot standby: two gateways share one identity; the platform gives the lease to one, only that one reports and executes writes, and the other takes over if it is lost.
- Several platform addresses with failover and automatic failback; WebSocket over port 443 for sites that allow only HTTPS; signed HTTPS upload of buffered data when the message bus stays unreachable (control commands still use the bus only).
- Redundant device paths: a second IP address, network card, serial server or standby CPU for the same device (Modbus TCP, SNMP, OPC UA), switched over after repeated failures and back on recovery, with every switch audited.
Outages and offline data packages
- During an outage readings are kept on local disk (by default up to 30 days or 2 GB) and backfilled in order; they are deleted only after the platform confirms them.
- For outages of several days an engineer exports a signed offline data package and imports it on the platform: de-duplicated, audited and acknowledged with a receipt signed by the platform.
Security
- Connects outward only: TCP 443 (HTTPS) and the gateway bus on TCP 4222 (TLS; not through a web proxy or CDN). No inbound port has to be opened at the site.
- TLS to the platform with one account per gateway; every command is Ed25519-signed and checked by the gateway before it runs.
- Installed with a one-line command and a one-time registration code; a gateway that goes offline raises a platform alarm.
Related
Control-room displays
Drag-and-drop dashboards for control-room screens: KPIs, trends, alarms, graphics and cameras, full screen in 1080p and 4K, with rotation across boards and sites.
Learn more →AI assistantRead-only operations assistant
Explains alarms, finds root causes, writes shift summaries and answers questions about site data, and turns statistical findings into maintenance suggestions. It never operates equipment.
Learn more →FMCSFacility management and control
The whole cooling and power chain at a glance, with redundancy, shift handover, utility efficiency and first-out alarm analysis.
Learn more →