Bandwidth

What is bandwidth?

Bandwidth is the maximum amount of data that a network connection, communication channel or transmission medium can carry per unit of time, expressed in bits per second (bit/s). In OT environments and industrial automation, bandwidth determines how much process data, alarms, video and management traffic can flow at the same time between PLCs, SCADA, HMIs, the historian, edge systems, IIoT devices and cloud platforms. That makes bandwidth a design criterion for reliability, scalability and availability. Yet high bandwidth says nothing about how quickly and predictably a single message arrives: latency and jitter measure that.

🧠 What exactly does bandwidth mean?

Bandwidth originally comes from signal theory, where it is the width of a frequency band in hertz. The Shannon-Hartley theorem (C = B · log₂(1 + S/N)) links that frequency width to the maximum data rate of a channel. In networking, the term has shifted to mean the nominal bit rate of a link, such as 100 Mbit/s for Fast Ethernet.

Bandwidth is measured in bits, not bytes. One byte is eight bits, so a 100 Mbit/s link carries at most 12.5 megabytes per second. Network vendors use decimal SI prefixes:

Bandwidth is a property of the link, but a chain is only as strong as its weakest link. A router or firewall performing deep packet inspection may handle less traffic than the ports connected to it.

🔄 What is the difference between bandwidth, throughput and latency?

These terms are often used interchangeably, but they measure different things. Bandwidth is the capacity of the pipe, throughput is what actually flows through it, and latency is how long a single message takes to arrive.

Throughput always stays below bandwidth. Besides its payload, every Ethernet frame carries a preamble (8 bytes), header (14 bytes), checksum (4 bytes) and a mandatory gap of 12 bytes. With a full 1,500-byte payload, roughly 95% of the line rate remains for TCP data. With small frames, such as cyclic I/O data, the relative overhead is much larger. A network with plenty of bandwidth can also still show high latency, for example because of full switch buffers or a long WAN link.

📊 How much bandwidth do industrial networks have?

Industrial networks combine media with very different capacities. Old serial links and fieldbuses remain in service for decades alongside modern Ethernet networks.

Wireless networks such as Wi-Fi, Bluetooth, LoRaWAN and 5G share their capacity among all users in a cell, and that capacity varies with distance and interference. That is why critical real-time connections are usually wired.

🏭 What traffic consumes bandwidth in an OT network?

Bandwidth plays a role in virtually all industrial processes. Traffic falls roughly into five streams:

• Cyclic process data - I/O data between PLC and field devices, in motion control sometimes every millisecond or faster. Small frames, but many per second.
• Supervision and history - SCADA polling, telemetry, historian replication and dashboards in Grafana fed from InfluxDB.
• Video and imaging - vision systems and camera surveillance. A single 1080p camera using H.264 compression typically needs several Mbit/s, continuously.
• Analytics and maintenance - vibration data for condition monitoring and predictive maintenance, often at high sampling rates, plus asset monitoring from the cloud.
• Management and security - logging, backups, firmware updates and security monitoring.

The choice of protocol affects how efficiently that bandwidth is used:

With Industry 4.0, cloud analytics and AI applications, it is mainly the traffic from the shop floor upwards through the Purdue Model that grows.

⚡ Why is more bandwidth not enough for real-time control?

Control requires predictable communication, not just plenty of capacity. A control loop expecting data every 2 ms runs into trouble when its frames repeatedly wait in a switch queue behind large video packets; after a few missed cycles the watchdog trips and the device goes into fault. A maximum-size Ethernet frame takes about 123 µs to transmit at 100 Mbit/s and about 12.3 µs at 1 Gbit/s, for every switch that stores and forwards it.

Industrial networks therefore combine bandwidth with mechanisms for deterministic behaviour:

• QoS - VLAN priority (IEEE 802.1Q) lets real-time frames overtake best-effort traffic.
• IRT - PROFINET IRT reserves time slots and synchronises clocks, enabling cycle times down to 31.25 µs with jitter below 1 µs.
• TSN - Time-Sensitive Networking brings scheduled time windows to standard Ethernet, so real-time and IT traffic can share one network.
• Tuning the cycle time - not every device needs to communicate every millisecond; doubling the cycle time halves the load.

IP can send I/O data via multicast. Without IGMP snooping on managed switches, that traffic reaches every port, including devices that do not need it.

🚨 What happens when bandwidth is insufficient?

When offered traffic exceeds the capacity of a link, congestion occurs. Switches buffer frames, buffers fill up and frames are dropped. Typical causes are broadcast storms from network loops, large historian dumps or backups during production, video streams on the process network, malware scanning the network and missing segmentation.

In cyber-physical systems and critical infrastructure, a network problem thus directly affects the physical process.

🔐 How are cybersecurity and bandwidth related?

Cybersecurity adds traffic of its own: logging to a SIEM, IDS sensors receiving a copy of the traffic, encryption with extra headers, and backups. Plan these flows too, and preferably route them over a separate management network.

Conversely, attacks can exhaust bandwidth:

IEC 62443-3-3 requires, through SR 7.1 (denial of service protection) and SR 7.2 (resource management), that a system keeps its essential functions under overload. Mitigating measures include network segmentation with VLANs and firewalls, rate limiting and broadcast storm control on switches, QoS and continuous monitoring. A sound security architecture therefore protects availability as well.

🔧 How do you plan bandwidth for an OT network?

Capacity planning starts with the traffic, not with the switch. A practical approach:

1. Inventory the flows - number of devices, protocols, cycle times, polling rates, cameras and historian tags per zone.
2. Calculate per link - add up the traffic crossing each uplink. In a line or ring topology, the uplink next to the controller carries the traffic of every device behind it.
3. Account for redundancy - redundancy with MRP or duplicate networks changes the traffic paths during a fault; the backup paths must handle the full load.
4. Keep headroom - size links so average utilisation stays low and peaks can be absorbed.
5. Separate and prioritise - put video, backups and IT traffic in their own VLANs and give real-time traffic priority.
6. Reduce data at the source - with edge computing and an edge device, you filter, aggregate and compress data before the WAN or cloud link, with local buffering during outages.

Worked example, serial bus: Modbus RTU at 9,600 bit/s uses 11 bits per character, so about 873 characters per second. Reading ten registers takes an 8-byte request and a 25-byte response, plus the mandatory 3.5-character silence between frames: just over 45 ms in total. Polling twenty slaves in sequence therefore takes almost a second, before any device response time.

Worked example, Ethernet: a small PROFINET frame occupies about 84 bytes on the wire including overhead. At a 1 ms update rate, that is roughly 0.7 Mbit/s per device per direction. A hundred devices behind a single 100 Mbit/s uplink already need about 67% of its capacity, before any other traffic.

📈 How do you monitor bandwidth?

Measure utilisation continuously, not only when something fails. SNMP reads interface counters from switches and routers, network monitoring tools turn them into graphs, and Wireshark analyses individual packets when there is a problem.

Mind the counters themselves: a 32-bit octet counter wraps in about 34 seconds at 1 Gbit/s. RFC 2863 therefore requires 64-bit octet counters (ifHCInOctets) on interfaces faster than 20 Mbit/s. A stable traffic baseline also helps detection: a sudden increase may point to a fault or an attack.

❓ Frequently asked questions

Is bandwidth the same as internet speed?

Bandwidth is the maximum capacity of a connection, while the speed you experience also depends on latency, overhead and how busy the network is. A 1 Gbit/s subscription delivers less throughput in practice. In OT networks, what matters most is whether messages arrive on time and predictably.

How much bandwidth does a PLC need?

A PLC with cyclic I/O communication typically uses less than 1 Mbit/s per device, because the frames are small. Total bandwidth depends on the number of devices, the cycle time and the topology. Video, historian replication and backups on the same network often need far more bandwidth than control itself.

Why is bandwidth measured in bits rather than bytes?

Bandwidth is expressed in bits per second because the physical layer transmits bits. File sizes are usually given in bytes, so divide a bandwidth by eight to estimate bytes per second. A 100 Mbit/s link thus carries at most 12.5 MB per second, before overhead.

Does more bandwidth solve latency problems?

More bandwidth shortens frame transmission time and reduces congestion, but it does not remove all latency. Delays caused by queuing, routing, firewalls or distance remain. Deterministic control needs QoS, TSN or PROFINET IRT in addition to sufficient bandwidth.

How much bandwidth does a wireless OT network have?

Bandwidth in wireless OT networks varies widely: WirelessHART reaches 250 kbit/s, LoRaWAN usually a few kbit/s, and Wi-Fi or 5G hundreds of Mbit/s. That capacity is shared by all devices in a cell and varies with interference. That is why fast, critical control loops are usually wired.

How do I measure bandwidth in an OT network?

You measure bandwidth usage through SNMP interface counters on managed switches, supplemented by flow data and packet analysis via a SPAN port or TAP. Prefer passive methods, because active speed tests themselves load a production network. Record a baseline so that deviations stand out.

📌 In summary

Bandwidth is the maximum capacity of a link in bits per second, but in OT what matters most is whether the right data arrives on time. Industrial networks combine fieldbuses running at kbit/s with Ethernet backbones running at Gbit/s. Successful OT convergence therefore requires capacity planning per link, segmentation, prioritisation of real-time traffic, data reduction at the edge and continuous monitoring, so that high availability and real-time control coexist with cloud, analytics and security.