The TCP/IP suite is how modern Ethernet-based networks — including most Industrial Automation and Control System (IACS) plant networks — actually move packets. Where the OSI Model is a teaching framework, TCP/IP is the practical stack: IP at the network layer, TCP or UDP at the transport layer, and application protocols (HTTP, Modbus TCP, OPC UA, and others) above.
Related: OSI Model | Switches and VLANs | Network Topologies | Modbus TCP | Network Segmentation | Network Attacks
Every device on a TCP/IP network needs a unique IP address, separate from the
Ethernet (MAC) physical address. IPv4 uses a 32-bit address written in
quad-dotted decimal (for example 192.168.10.15), allowing roughly 4.3 billion
addresses.
Subnet membership is defined by a subnet mask:
255.255.255.0255.255.255.0
equals /24On a local Ethernet segment, IP packets must still be delivered to a MAC address. Address Resolution Protocol (ARP) asks the LAN: “Who has this IP?” The owner replies with its MAC. Devices cache recent IP↔MAC mappings in an ARP table to reduce broadcasts.
ARP is a frequent target for local attacks (see ARP spoofing). IPv6 does not use ARP; it uses neighbor solicitation instead.
Routers (Layer 3) forward packets between networks using static or dynamic routing tables. Hosts use the same ideas:
ICMP (Internet Control Message Protocol) carries diagnostics and error reports (for example unreachable destinations). Engineers use it for troubleshooting; attackers and scanners also probe with it — see network discovery and scanning.
A classic router connects LANs and often a WAN, with an IP routing table. A Layer 3 switch can switch within a LAN like Layer 2 and also route between subnets — typically without a WAN edge. Collapsing accounting, engineering and PLC networks onto one L3 switch can cut cost and increase flexibility, but concentrates risk: evaluate segmentation and trust boundaries carefully (network segmentation).
The transport layer makes end hosts appear directly connected: numbering, reordering, end-to-end error recovery and flow control as needed.
| Protocol | Style | Typical use |
|---|---|---|
| TCP | Connection-oriented; reliable delivery, acknowledgements, retransmission | Sessions that must not lose data (many IT apps; some industrial protocols) |
| UDP | Connectionless; low overhead, no guaranteed delivery | Fast streams, discovery, some real-time or simple industrial traffic |
TCP and UDP identify applications with port numbers. The combination of source and destination IP + port steers data to the correct service.
| Range | Name | Notes |
|---|---|---|
| 0–1023 | System / well-known | e.g. HTTP 80, DNS 53, many industrial defaults |
| 1024–49151 | User / registered | Registered application ports |
| 49152–65535 | Dynamic / ephemeral / private | IANA suggested client return ports; OS defaults vary (older Windows used narrower ranges; many Linux builds use other ephemeral ranges) |
Servers typically listen on well-known or registered ports; clients often use ephemeral return ports assigned by the OS.