From Pipeline SCADA to 15 Million Tags: MQTT, Sparkplug & the Unified Namespace


Arlen Nipper co-invented MQTT in 1999 to connect an oil pipeline across a satellite link. Twenty-five years later it underpins 15 million live tags at a 3-second SLA — and the architect says he is still not surprised.
In his third appearance on Unplugged, Arlen joins Phil and Ed to map the full arc of industrial connectivity: from cryptic PLC registers with no engineering units to unified namespaces powering AI training in real time. He covers data centres evolving into AI factories, renewable energy consolidating hundreds of remote sites into a single operations centre, and what Sparkplug v4 will address in alarm pipelines.
Along the way he explains why the Unified Namespace market grew from zero adoption to more than $2 billion, why data centre customers are standardising on MQTT 3.1.1 over MQTT 5, and where AI actually fits in the OT stack — hint: not until you have a proper single source of truth.
15 million tags, 3-second SLA. Arlen described an active deployment where 15 million tags flow continuously — many changing every second — with a contractual 3-second guarantee from field change to the Ignition platform. The mechanism is report-by-exception: “That’s really the only way you’re going to get the bandwidth that you require.”
Remote Operations Centres. NextEra Energy took 524 solar, BESS and wind sites and consolidated control into a single Remote Operations Centre — site to OT broker to Ignition to IT broker to cloud — creating a new profit centre by routing energy efficiently across generation and storage assets.
Data centres as AI factories. What started as accretive infrastructure — HVAC, power and cooling bolted on separately — is now an integrated AI factory: millions of Modbus sensors at the rack level, ingested via Ignition Edge, then MQTT/Sparkplug to the broker. “I haven’t had one data centre customer yet that didn’t come to us and say, ‘okay, we’re gonna use MQTT.’”
AI and the Unified Namespace. Arlen’s framework for connecting AI to OT data: decouple, establish a single source of truth, and only then apply AI. Define the process in a UDT, publish via Sparkplug, train AI on the live structure, and republish when a process variable changes so the enterprise always holds the current truth.
MQTT 3.1.1 vs MQTT 5. A large data centre customer wrote a white paper rejecting MQTT 5 as “a lot of fluff and overhead” and committed entirely to 3.1.1. Arlen’s take: Sparkplug dominates IIoT and MQTT 5 has not broken through in this audience.
Arlen Nipper — President & CTO, Cirrus Link Solutions. Arlen co-invented MQTT in 1999 with Andy Stanford-Clark while working to connect an Amoco oil pipeline across a satellite network. He is a key author of the Sparkplug specification and has more than 47 years of experience in SCADA and industrial automation. In 2025 he received the inaugural Lifetime ProveIt! Award.
Previous appearances: EP1 and EP24.