A 100 Mbps industrial switch is not obsolete — it is outgrown. It still carries PLC polling and sensor readings with plenty of margin, and it stops being adequate the moment a vision system, a historian or an MES integration is added to the same cell. Rayin (Shenzhen Rayin Technology) is a manufacturer of GPON OLTs and industrial Ethernet switches. The upgrade decision is not about speed for its own sake; it is about which traffic class you added last.

Fast Ethernet fails on burst, not on average. PLC traffic is steady and small; machine vision, firmware transfers and historians are neither.
A single uncompressed machine-vision camera can exceed a 100 Mbps access link on its own — a 5-megapixel monochrome camera at 20 frames per second is roughly 800 Mbps before compression.
The uplink dimensioning matters more than the access ports. Twenty-four gigabit access ports sharing one gigabit uplink is a 24:1 oversubscription by design.
The pattern is consistent across plants: the control traffic that justified the original design stays small, and everything added afterwards is large.
| Traffic class | Typical demand | Fits on 100 Mbps? |
|---|---|---|
| PLC polling, sensor values | Kilobits to low megabits | Comfortably |
| HMI and SCADA polling | Low megabits | Yes |
| Compressed HD camera | 4–8 Mbps | Yes, one or two streams |
| Uncompressed vision camera | Hundreds of Mbps | No |
| Historian bulk upload | Bursty, tens to hundreds of Mbps | No, it bursts |
| Firmware and image distribution | Large one-off transfers | Slow, and disruptive |
| AGV fleet telemetry and maps | Tens of Mbps, latency-sensitive | Marginal |
The failure mode is instructive. Average utilisation on a 100 Mbps link may sit under 20% while the cell still drops frames, because the historian upload or the vision burst fills the port buffer for the few hundred milliseconds that matter. Average utilisation is the wrong metric; look at microbursts and interface discards.
The decision splits by layer, and mixing them is normal.
| Situation | Right choice | Reasoning |
|---|---|---|
| Sensors and PLCs only, no new traffic | Keep the existing ports at the edge | No benefit in replacing working ports |
| Vision, MES or historians added | Gigabit copper access | Burst headroom on every port |
| Many gigabit ports aggregating | 10G fibre uplink | Removes the aggregation bottleneck |
| Long runs between buildings | Fibre, not copper | Copper is capped at 100 m per link |
An 8+4-port gigabit industrial switch fits the middle case well: eight copper ports for cell devices, four SFP slots giving a fibre uplink plus a ring path. Where several such switches aggregate, a 10G uplink is what stops the uplinks from becoming the bottleneck the access upgrade just removed.
Two checks before you order:
Count the aggregation ratio. If the sum of active access ports exceeds the uplink by more than you are comfortable oversubscribing, the uplink needs to be ten times faster, not equal.
Check what the device supports. Many field devices are still 100 Mbps. Gigabit ports auto-negotiate down to them, so replacing the switch is safe — but a 100 Mbps device stays a 100 Mbps device no matter what it plugs into.
More ports and more speed are the headline, but four features decide whether the upgrade actually helps:
| Feature | Why it matters in a mixed-speed cell |
|---|---|
| QoS with DSCP or 802.1p priority | Keeps control traffic ahead of bulk transfers during a burst |
| IGMP snooping | Stops multicast video flooding every port on the switch |
| Jumbo frames | Reduces interrupt load on high-throughput endpoints — enable only where every device agrees |
| Ring or dual-uplink protection | Recovery in tens of milliseconds instead of a spanning-tree reconvergence |
Management is the fifth item and the easiest to forget. A switch you cannot poll, back up or audit will be configured once and then left alone, which is how a good upgrade turns into an undocumented one. Confirm SNMP, a CLI or a usable web interface, and a configuration export that can be restored onto a replacement unit.
Phased replacement works because the ports are backward-compatible:
Measure first. Record interface discards and microburst behaviour on the existing switch for a full production week before deciding anything.
Replace the aggregation point first. Upgrading the uplink before the access ports often resolves the symptom on its own.
Swap cell switches during planned stops. Keep the old switch as a warm spare; a retired access switch is a cheap insurance policy.
Re-measure after. If discards have not dropped, the constraint was never the access speed — check the uplink, the server NIC or the storage target instead.
Is a 100 Mbps industrial switch obsolete? No. For cells carrying only PLC and sensor traffic it remains adequate. It becomes a constraint when vision, historians or bulk transfers share the same port.
Why does the network drop frames when utilisation looks low? Because the metric is wrong. Short bursts fill the port buffer even when the one-minute average is low, so check microbursts and interface discards rather than average utilisation.
Do I need 10G at the access port? Rarely. 10G belongs on uplinks and aggregation points; the access ports need gigabit headroom, which is a tenfold increase on its own.
Will a gigabit switch work with my 100 Mbps field devices? Yes. Gigabit ports auto-negotiate down to Fast Ethernet, so mixed-speed cells work — the device stays at its own speed.
What is the cheapest first step? Measure, then upgrade the uplink. Many plants find the aggregation point was the bottleneck and the access switches were never the problem.
About the author: Sara is a Customer Manager at Rayin with over 10 years of experience in the communications field. In her free time, she enjoys badminton and swimming.
About Rayin: Shenzhen Rayin Technology Co., Ltd. — company profile