LADX
Start free
← Resources
Networking·7 min read·

Choosing a switch for an industrial network

Managed against unmanaged, the features that matter for PROFINET and EtherNet/IP, and why the cheap one in the office cupboard causes trouble.

Short answer

For PROFINET, yes in practice. A managed switch gives per port diagnostics, prioritisation of real time traffic, IGMP snooping to contain multicast, and participation in LLDP so the topology is visible. Unmanaged switches pass frames and report nothing, which turns every intermittent fault into a process of elimination by unplugging things.

Which protocol sits whereENTERPRISEOPC UAMQTTCONTROLEtherNet/IPPROFINETModbus TCPDEVICEIO-LinkPROFIBUS DPModbus RTUEthernet carries most of it now, which is why the layers blur.

A switch is a switch until the network starts misbehaving intermittently, at which point the difference between a managed and an unmanaged one is the difference between an afternoon and a fortnight.

What managed actually buys

Per port diagnostics. Counters for frames discarded, CRC errors, collisions and link transitions. This is the single most valuable feature and the reason to accept the cost.

Prioritisation. Real time protocol traffic marked and given priority over a firmware download or a camera stream sharing the same link.

IGMP snooping. Multicast delivered only to ports that asked for it, rather than flooded everywhere.

LLDP. Neighbour discovery, so the engineering tool can draw the actual topology and compare it with the intended one. On PROFINET this also underpins device replacement without reconfiguration.

Ring redundancy. MRP or a vendor equivalent, so a single cable break does not stop the line.

The multicast problem, specifically

Worth a paragraph because it bites hard on EtherNet/IP.

Implicit I/O connections use multicast. A switch that does not do IGMP snooping treats multicast like broadcast and floods it to every port.

On a small network with a handful of devices, this is invisible. On a network with forty drives, every drive receives every other drive's I/O traffic and has to discard it in software. The symptom is devices dropping under load with no obvious cause, and the fix is a switch feature rather than anything in the program.

Why office hardware fails in a panel

The environmental differences are not marginal.

  • Temperature. A panel at 50 degrees internal is normal. Office switches are rated to about 40 and derate before that.
  • Power. 24 volt DC, often from the same supply as everything else, with the dips that implies. An office switch expects a mains adaptor.
  • Vibration. RJ45 latches work loose. Industrial switches use screw retained connectors or M12 for a reason.
  • Mounting and life. DIN rail, convection cooling, no fan. A fan in a panel is a filter to clean and a bearing to fail.
The failure mode is the problem, not the failure. An office switch in a panel does not die cleanly; it starts corrupting frames intermittently when warm, which presents as a network problem for weeks.

Segmentation

One flat network for the whole plant is simple and becomes the thing that makes an outage plant wide.

The usual arrangement is a separate network per machine or cell, joined to a plant network through a router or a firewall, with the machine network able to run when the plant network cannot.

This also contains a broadcast storm and gives a clean boundary for the security controls that will eventually be required.

What is genuinely enough

Not every network needs a top of the range switch.

  • A machine with a handful of devices and no motion: a small managed switch with diagnostics and QoS.
  • A line with EtherNet/IP drives: managed with IGMP snooping, not negotiable.
  • A PROFINET machine with a ring: managed with MRP and LLDP.
  • A plant backbone: fully managed, with VLANs and a maintained configuration.

The one to avoid at every level is the unmanaged five port unit that solved a problem quickly and is now load bearing.

Common questions

Why does EtherNet/IP need IGMP snooping?
Because implicit I/O messaging uses multicast. Without snooping, a switch floods multicast to every port, so every device receives every other device's I/O traffic. On a small network it is survivable; on a large one it saturates links and overloads device processors.
Can I use an office switch on a machine?
It will pass frames, and it will fail early. Office switches expect still air, mild temperatures, no vibration and mains power. In a panel at 50 degrees on a vibrating machine, with 24 volt supply, they fail in ways that look like network faults for weeks before anybody suspects the switch.
What is the practical difference for troubleshooting?
A managed switch tells you which port is discarding frames and which cable has CRC errors, so an intermittent fault is located in minutes. With unmanaged switches the only diagnostic technique available is unplugging things until it changes.

Keep reading