Industries / Industrial equipment and services / IoT and industrial automation / SCADA integration
Upgrades
This is the work of replacing a SCADA system's computers, software and links before they lose support, and of changing a running system without losing sight of the plant.
4
Milestones achieved, from 2008 to 2022.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
2008
An update trips a reactor
On 7 March 2008 Unit 2 of the Hatch nuclear plant near Baxley, Georgia, shut down automatically, and stayed down for 48 hours, after a Southern Company engineer installed a software update on a computer on the plant's business network. That computer exchanged data with a control system; when it rebooted, the control system's data reset, and safety systems read the missing data as a drop in the water that cools the fuel, the Washington Post reported from the operator's filing with the Nuclear Regulatory Commission. The engineer had not known the two machines synchronized.
LessonBefore an update, list every system the machine trades data with, including links across the boundary between business and control networks.
Credit
Southern Company (operator); Brian Krebs, Washington Post (report)
2013
A bad edit copies itself
NERC, the body that sets reliability rules for North America's grid, described a utility whose energy management system (EMS), the SCADA system of a grid control centre, failed during a database update. A server restart during the attempt to undo the edits created faulty files that synchronized across the integrated servers, and all three communications servers failed. The update had been tried first on development and training systems, but their error logs were only partly reviewed.
LessonRedundant servers copy a bad change as faithfully as a good one.
Credit
NERC (lessons learned, 29 October 2013)
2014
Windows XP support ends
Microsoft stopped security updates for Windows XP on 8 April 2014, having announced the date in 2007. IHS, an analysis firm, wrote that XP was the most widely used operating system in industrial automation, running industrial PCs and control systems, and that larger companies had upgraded early while many smaller ones had not.
LessonThe software under a SCADA system has a support end date set by its maker, and plants must plan replacements around it.
Credit
Microsoft (end of support); IHS (assessment)
2022
3G networks switch off
US carriers closed their 3G networks to free radio spectrum for newer service: AT&T by February 2022, T-Mobile's in stages from March to July 2022 and Verizon by 31 December 2022, the FCC said. The International Society of Automation's water and wastewater division warned that cellular modems carrying SCADA data for power distribution, gas utilities, pipelines and water systems had to be replaced with 4G/LTE units. AT&T had already retired its 2G network in 2017.
LessonA link on a public network ends on the carrier's timetable, not the utility's.
Credit
AT&T, T-Mobile and Verizon (shutdowns); FCC (schedule)
Watch for
AI migrates old screens
Tatsoft says FrameworX AI, generally available since May 2026, lets an AI client such as Claude read and write a SCADA project's tags, alarms and screens through MCP (Model Context Protocol) servers, the open standard by which AI clients call software tools. It estimates that migrating 40 legacy displays with their tags, alarms and scripts takes "hours to a few days" with the tool, against 160 to 320 hours by hand, and says a conference exercise budgeted at 72 hours of engineering took 2.
Achieved whenan outside party migrates the same legacy project with and without the tool and publishes the hours spent and the errors found, screen by screen.
Claimed by
Tatsoft, May 2026
Copper phone lines close
Openreach says all traditional phone lines in the UK will go digital on 31 January 2027. Remote pumping stations and other unattended sites still send alarms and telemetry over analogue lines; a guide from CSL, which sells replacements for them, cites UK government guidance counting more than 2,600 major phone network incidents in 2024/25, including exchange failures that hit alarm and telemetry systems.
Achieved whenOfcom, the UK communications regulator, confirms the analogue network has been withdrawn.
Claimed by
Openreach, 2026
Open
01
Clocks past 2038
A public bench test of RTUs, PLCs, operator panels and SCADA servers whose clocks are run past 03:14:07 UTC on 19 January 2038, the moment a signed 32-bit count of seconds since 1970 runs out (2,147,483,647 seconds), recording what each one does.
The briefCredit · unclaimed02
Check migrations point by point
An open tool that reads the old and the new SCADA system side by side during a parallel run and lists every point whose value, status, units or alarm differs, so mapping errors in a migration are found before the old system is switched off.
The briefCredit · unclaimed
Open 01 · the brief
Clocks past 2038
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public bench test of RTUs, PLCs, operator panels and SCADA servers whose clocks are run past 03:14:07 UTC on 19 January 2038, the moment a signed 32-bit count of seconds since 1970 runs out (2,147,483,647 seconds), recording what each one does.
Why it is still open
Makers are finding the fault one product line at a time: CERT@VDE's advisory VDE-2025-007 of 15 April 2025 (CVE-2025-0101) says a user can set some WAGO controllers to 19 January 2038, after which some functions "work unexpected or stop working at all." Research on the problem in energy systems simulates it: Saeid Jamshidi and colleagues added "epoch-overflow events" to recorded energy IoT data to train a detector of timing faults. Searches of GitHub and arXiv found no published cross-maker results from real control devices run past the date.
Built with what exists
Surplus RTUs, PLCs and panels bought second-hand; a time server you control, such as chrony, serving a future date on an isolated network; Wireshark; the trial edition of a SCADA package to log what the control room would see.
First step
Set one surplus RTU and its master to 18 January 2038 on an isolated bench and record the next 48 hours.
Done looks like
At least 30 device and software models from at least eight makers run across the date on an isolated bench, with a published table of what each did (wrong dates, stopped logging, restarts, rejected time) and the test procedure.
Open 02 · the brief
Check migrations point by point
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open tool that reads the old and the new SCADA system side by side during a parallel run and lists every point whose value, status, units or alarm differs, so mapping errors in a migration are found before the old system is switched off.
Why it is still open
NERC's December 2021 analysis of 2018–2020 control centre outages names management and organization, including change management, as the leading root cause, and found that both losses of automatic generation control, the software that keeps supply matched to demand, came from bugs introduced in routine software updates; its 2013 case shows a change tested in advance whose error logs were only partly read. Commercial simulators such as Triangle MicroWorks' Distributed Test Manager automate communication tests in substations, but no open tool for comparing two live systems point by point, and no published count of the errors such tools catch, turned up on GitHub.
Built with what exists
Open-source protocol libraries such as Step Function I/O's dnp3, pymodbus and asyncua (for OPC UA); two SCADA packages with free or trial editions, such as FUXA and Ignition; a time-series database such as InfluxDB to line up readings from both systems.
First step
Map 50 points from a simulated DNP3 outstation into two free SCADA packages, plant five errors, and see which a short comparison script finds.
Done looks like
On a bench of 1,000 points mirrored into two systems, with 50 mapping errors planted at random (wrong index, wrong scaling, inverted status, missing alarm), the tool finds at least 48 with no more than five false flags, and the bench setup and planted list are published.