Stop Chasing the Newest Protocol. The Old Ones Still Win.

Every few months someone tells me the industrial plant floor is about to be swept clean by whatever protocol just won a trade-show award. I don't buy it. The most useful thing you can do for your network this year is not adopt TSN, not rip out PROFIBUS, and not bet your architecture on whichever fieldbus has the loudest marketing budget. It's to get very, very good at the protocols you already own.

I'll defend that against the obvious objection in a moment. First, the thesis: in industrial networking, protocol longevity beats protocol novelty, and the protocols worth your engineering hours are the ones with the deepest installed base—not the ones with the best spec sheets.

The installed base is the real standard

Modbus was developed in the late 1970s and still uses the same master-slave register model it always did, over serial RTU or Ethernet TCP. The Modbus Organization flatly calls itself the steward of the world's most widely adopted industrial communication protocol, and I have no reason to argue. You can walk into almost any plant built in the last forty years and find a Modbus device somewhere doing something important. That's not nostalgia; that's an asset base.

The same logic scales up. PROFIBUS & PROFINET International reports more than 68 million PROFIBUS nodes, over 69 million PROFINET nodes, and over 51 million IO-Link nodes installed worldwide as of the end of 2023 (PROFIBUS & PROFINET International). Those are not aspirational numbers from a roadmap slide. They are machines that exist, are maintained, and will be specified again on next year's projects because the spare parts, the engineers, and the training already exist.

EtherNet/IP tells the same story in North America, running the Common Industrial Protocol over standard, unmodified Ethernet on Rockwell and Allen-Bradley platforms (ODVA EtherNet/IP). CIP itself is media-independent, object-oriented, and supported by hundreds of vendors, with native translation services for Modbus, HART, and IO-Link devices (ODVA Common Industrial Protocol). That translation layer matters more than any new wire-level feature. It means your brownfield mess doesn't have to be a religious war.

What the newcomers actually change

None of this means newer protocols are pointless. It means they win in specific places, not everywhere. EtherCAT is genuinely excellent at hard real-time motion: cycle times down to 100 microseconds or less, jitter under a microsecond, and cable redundancy that recovers from a break in under 15 microseconds so a motion application loses at most a single cycle (EtherCAT Technology Group). Sercos, standardized under IEC 61784 and IEC 61158, pushes configurable cycle times from 31.25 microseconds to 65 milliseconds with synchronization accuracy well under a microsecond, and recognizes a ring cable break within 25 microseconds (Sercos International). If you are building a printing press or a packaging line with tight axis synchronization, these are the right tools. If you are building a tank farm, they are almost certainly overkill.

Above the field level, the split is even cleaner. Protocols are layered: Modbus, PROFINET, and EtherNet/IP connect devices and PLCs, while OPC UA and MQTT carry data securely up to SCADA, MES, and cloud systems (OPC Foundation). OPC UA is standardized as the IEC 62541 series and is vendor-neutral and platform-independent with built-in encryption and authentication. MQTT is standardized as ISO/IEC 20922 and is deliberately lightweight—tiny clients, small headers, publish/subscribe decoupling, and three quality-of-service levels running from at-most-once to exactly-once delivery (MQTT.org). These two are not competing with your fieldbus. They are doing a different job.

And IO-Link deserves a special mention because it is routinely misunderstood. It is not a fieldbus at all. It is the first globally standardized I/O technology, IEC 61131-9, running point-to-point over the ordinary three-wire sensor cable you already pulled (IO-Link Consortium). It is the cheapest digital upgrade most plants can make, and it slots under PROFINET or EtherNet/IP without argument.

The counter-argument, and why I still say no

The strongest objection is security, and it is a good one. CISA warns that many legacy ICS devices still run outdated operating systems and older protocols that lack encryption or authentication, and it singles out brownfield deployments—layering modern OT on top of legacy infrastructure—as a challenge unique to ICS. CISA's own Secure by Demand guidance notes that threat actors exploit weak authentication, insecure settings, and outdated protocols, and recommends buying products with open standards, strong authentication, and data protection.

I take that seriously, and I still reject the conclusion that the answer is protocol replacement. The fix for an insecure Modbus segment is not to tear it out; it is to segment it, gateway it, authenticate at the boundary, and stop exposing it. NIST SP 800-82 Rev. 3 is explicit that OT security is about protecting systems that interact with the physical world, and ISA/IEC 62443 gives you the requirements framework across asset owners, suppliers, integrators, and service providers. Note that EtherNet/IP already offers CIP Security, which is designed to help users comply with IEC 62443 and the EU CRA without abandoning the protocol. You defend the installed base; you don't replace it on a security scare.

Here's the concrete version. Picture a mid-size water treatment site with a 1990s Modbus RTU loop on the clarifier, a PROFINET segment on the newer pump skids, and an EtherNet/IP line on the bottling side. The tempting move is to standardize everything on one modern stack. The right move is cheaper and faster: leave the Modbus loop alone, drop an IO-Link master onto the instrument air skid so you get sensor diagnostics you never had, and put an OPC UA server at the cell boundary so the SCADA and MES layers get encrypted, authenticated data without anyone touching the PLC logic. You just improved visibility and security without a single production outage.

Where the money actually is

The market data backs the boring view. Hardware—robots, PLCs, sensors, HMI panels—still holds roughly 50% to 60% of the industrial automation market (Maximize Market Research), and hardware is the largest segment of the IIoT market ahead of software and platforms. The global industrial automation market was about $184.43 billion in 2025 and is projected to reach $326.48 billion by 2032 (Maximize Market Research). Asia Pacific holds the largest regional share at roughly 35% to 42%, and MarketsandMarkets projects the region's industrial control and factory automation market growing from $95.73 billion in 2025 to $194.52 billion by 2032. All of that hardware ships with protocol interfaces already decided. Your job is to connect it, not to relitigate it.

So here is my recommendation, stated plainly: build protocol competence, not protocol fashion. Pick the fieldbus your vendors and your maintenance techs already know, use IO-Link to digitize the last few meters, and put OPC UA and MQTT at the boundary where they belong. Treat TSN, EtherCAT, and Sercos as targeted instruments for hard real-time work, not as the new default. And treat your legacy protocols as infrastructure to be secured, not sins to be confessed.

The single most important thing to remember: in industrial networking, the protocol that already runs your plant will outlive the one on the trade-show banner—so invest in mastering and securing what you have, and adopt new protocols only where they solve a problem the old ones genuinely cannot.

Sources

• OPC Foundation - https://opcfoundation.org/
• PROFIBUS & PROFINET International (PI) - https://www.profibus.com/technology/
• ODVA EtherNet/IP - https://www.odva.org/technology-standards/key-technologies/EtherNet-ip/
• CISA Industrial Control Systems - https://www.cisa.gov/topics/industrial-control-systems
• Maximize Market Research - https://www.maximizemarketresearch.com/
• IO-Link Consortium - https://io-link.com/