Utility connectivity has a distinctive shape: hundreds of unmanned sites, most of them in the wrong place for signal, running SCADA traffic that hardly touches the bandwidth but punishes every lost packet. The engineering answer is not to hunt for one flawless link per substation, because there isn't one; it is to build a logical network that treats every imperfect link as raw material for a reliable one.
Why the legacy approach fails
Utility WANs were largely built on MPLS circuits and single-path cellular, which was fine for low-frequency telemetry and is not fine now. The failure modes are familiar to anyone running this estate: a cable strike during civil works darkens a site for days; a congested or fading cell path drops just long enough for SCADA sessions to time out, leaving data gaps that are painful to reconcile; and standard failover switches only after the primary has fully died, which for stateful control traffic is already too late.
The regulatory backdrop raises the stakes. Continuous logging is increasingly a compliance requirement, so a connectivity gap is not just an operational nuisance, it is a monitoring gap with potential fines and safety implications attached. And because more of this traffic now crosses public networks, encryption in transit has to be designed in from the start rather than bolted on with a performance-sapping VPN afterthought.
Software-defined resilience
SpeedFusion changes the redundancy model. Instead of a backup link idling until the primary dies, every available path, 4G, 5G, LEO satellite, fixed line where one exists, carries traffic continuously inside a single encrypted tunnel. Packet-level bonding means a link failure changes the capacity of the tunnel, not the state of the session; the RTU stays connected, the historian keeps logging, and nobody reconciles a gap.
For the SCADA traffic itself, the tool that earns its keep is WAN Smoothing: critical packets duplicated across paths, first arrival wins, duplicates discarded. Telemetry is low-bandwidth, so the duplication overhead is cheap insurance against the false alarms and lost RTU synchronisation that single lost packets cause. We prioritise control-plane traffic explicitly, so a firmware download or CCTV stream can never elbow a control signal out of the way.
Substations, wind farms and the unstaffed estate
The estate problem is as important as the link problem. A distribution network operator's sites number in the hundreds, unstaffed, dispersed, and expensive to visit. InControl2 makes that estate operable from one place: link health and signal per site, remote configuration, staged firmware rollouts, and per-SIM data tracking that catches an anomalous consumer before the bill does. LEO satellite has quietly transformed the worst locations; a moorland substation or offshore-adjacent wind asset that once justified a leased line can now bond satellite with whatever cellular exists, and hold a dependable tunnel where neither path alone would.
Legacy equipment integrates rather than blocks: serial-attached RTUs and older PLCs bridge through the same gateway into the encrypted tunnel, so decades-old assets appear in modern monitoring without replacement programmes.
Visibility beyond the network team
In utilities, the people who need to know a site's status are often not network engineers: operations controllers, compliance officers, field supervisors. We build portals on top of the Peplink estate that translate link telemetry into operational language, this site healthy, that site degraded, this one running on satellite since Tuesday, and feed the same data into existing enterprise monitoring where it belongs. The network stops being a black box owned by one team and becomes an instrumented part of the infrastructure.
Deployment as an engineering exercise
We scope utility networks from the application inwards: SCADA protocol behaviour and latency tolerance, compliance logging requirements, realistic coverage at each site class, then the transport mix, hardware and tunnel design to serve it. Rollouts across hundreds of sites live or die on repeatability, so we build configuration templates per site class, stage them through InControl2, and train the internal team to own the estate day to day, with our managed service behind them for the faults that would otherwise mean a long drive to a locked compound.
The short version
Bond every available path into one encrypted tunnel, smooth the SCADA traffic so single packets stop mattering, manage the whole estate from one screen, give operations a view they can read, and template everything so site five hundred deploys like site five. That is utility connectivity that survives contact with the real world. If your remote sites are still hostage to single links, get in touch for a scoping conversation.
Frequently asked questions
Our SCADA traffic is tiny. Why does connectivity design matter so much?
Because SCADA punishes loss and latency, not bandwidth shortage. A single dropped packet can false-alarm a monitoring system or desynchronise an RTU, so the design goal is packet delivery consistency, which is exactly what smoothing and bonding provide.
Can we keep our existing MPLS while migrating?
Yes, and it is often the sensible path: the MPLS circuit joins the bonded tunnel as one path among several, resilience improves immediately, and the circuit can be retired on its own timetable rather than as a big-bang cutover.
Is satellite viable for compliance-grade monitoring?
Bonded with cellular, yes. LEO latency is workable for telemetry, and the bond covers each path's weaknesses. As a sole link it inherits single-provider risk, which compliance-grade monitoring should not accept.
How do you handle sites with serial-only legacy equipment?
The Peplink gateway bridges RS-232/RS-485 into the IP network and carries it inside the encrypted tunnel, so legacy RTUs and PLCs join centralised monitoring without hardware replacement.
What does managing hundreds of sites actually look like day to day?
Mostly quiet, if built properly: template-driven configuration, staged firmware waves, per-site health on one dashboard, and alerts tuned so a human only looks when something genuinely needs a decision.