14.3.3 Ethernet Switching And Logical Segmentation

While the Ethernet frame defines how data is formatted and addressed at the Data Link Layer, it does not determine how frames are forwarded within a local network. In modern deployments, that function is performed by Ethernet switches operating at Layer 2. Switches interconnect devices using point-to-point links arranged in a physical star topology, replacing the shared collision domains of early Ethernet with independent full-duplex links and enabling simultaneous transmission across multiple ports. We therefore next examine switching operation and frame forwarding in contemporary LAN architectures.

14.3.3.1 Basic Switching Operation

An Ethernet switch operates primarily at the Data Link Layer (Layer 2 of the OSI model). Its principal function is to forward frames based on destination MAC addresses. When a frame arrives at a switch port:

The switch reads the source MAC address and records it in a forwarding table along with the ingress port.
The switch examines the destination MAC address.
If the destination address is known, the frame is forwarded only to the corresponding port.
If the destination is unknown, the frame is flooded to all ports except the ingress port.

This process is known as MAC learning and allows the switch to dynamically build a forwarding table without manual configuration.

14.3.3.2 Collision Domains And Broadcast Domains

Ethernet switching fundamentally changed how contention and frame propagation occur within a local network. To understand this transition, it is necessary to distinguish clearly between collision domains and broadcast domains, as they represent different aspects of network behavior at Layer 2.

Collision Domains

In early shared-medium Ethernet, all devices attached to the same bus formed a single collision domain—that is, a portion of the network in which frames transmitted by one device could collide with frames transmitted by another. In legacy coaxial Ethernet, the entire segment constituted one collision domain, and repeaters extended that same domain to the maximum permitted network diameter. Access to the shared medium was therefore regulated using CSMA/CD.

In modern switched Ethernet, each device connects to a switch port via a dedicated point-to-point link operating in full-duplex mode. Because transmission and reception occur on separate signal paths and no shared medium exists, collisions cannot occur.

Accordingly, each switch port forms an independent collision domain, but in full-duplex operation the concept has no practical significance. CSMA/CD is disabled, and collision-domain constraints no longer influence network design.

Full-Duplex Operation

The elimination of collisions in modern Ethernet is a direct consequence of full-duplex operation. In historical half-duplex Ethernet, transmission and reception shared the same physical medium, so a device could either transmit or receive at a given time, but not both. Because multiple devices contended for access to the shared medium, collision detection using CSMA/CD was required.

In full-duplex Ethernet, separate signal pairs (for copper) or separate fibers (for optical links) are used for transmission and reception. A device may therefore transmit and receive simultaneously, and no shared contention medium exists. As a result, collisions cannot occur and CSMA/CD is disabled. Effective throughput is doubled relative to half-duplex operation at the same nominal data rate.

Full-duplex capability is negotiated automatically during link establishment. In practice, virtually all contemporary Ethernet deployments operate in full-duplex mode.

Broadcast Domains

While switching and full-duplex operation eliminate collisions, they do not eliminate broadcast propagation. A broadcast domain is defined as the set of devices that receive broadcast frames transmitted by any member of that set.

Ethernet specifies a broadcast MAC address (FF:FF:FF:FF:FF:FF). Frames addressed to this value are delivered to all devices within the same broadcast domain. In a basic switched LAN without VLAN configuration, all switch ports belong to a single broadcast domain. When a broadcast frame is received, the switch floods it to every port except the port on which it was received. As the number of connected devices increases, excessive broadcast traffic can consume bandwidth and reduce overall network efficiency.

Unlike collision domains—which are effectively eliminated in modern full-duplex Ethernet—broadcast domains remain a significant architectural consideration. Managing broadcast scope therefore becomes an important aspect of scalable LAN design.

VLANs: Logical Segmentation of Broadcast Domains

Logical segmentation of broadcast domains in switched Ethernet is achieved using Virtual Local Area Networks (VLANs), standardized in IEEE 802.1Q. A VLAN allows a switch to partition its ports into separate logical networks, even though they share the same physical switching infrastructure.

IEEE 802.1Q defines a VLAN tag that is inserted into the Ethernet frame, identifying the logical network to which the frame belongs. Switches use this VLAN identifier to ensure that frames are forwarded only to ports assigned to the same VLAN. Broadcast, multicast, and unknown unicast traffic is therefore confined to the appropriate logical group.

Devices assigned to different VLANs cannot exchange Layer 2 frames directly and do not receive each other’s broadcast traffic. Communication between VLANs therefore requires routing at Layer 3, typically implemented by a router or a Layer 3 switch.

By dividing a physical switch into multiple logical segments, VLANs reduce broadcast domain size, improve administrative flexibility, and enhance network security through traffic isolation. Each VLAN constitutes an independent broadcast domain, even though multiple VLANs may coexist on the same switch hardware.

Without VLAN configuration, all switch ports within a LAN would belong to a single broadcast domain. VLANs therefore provide the principal mechanism for controlling broadcast scope within modern switched networks.

It is important to note that VLANs provide segmentation within a local switched network but do not themselves create secure connectivity across multiple routed networks or geographically separated sites. The extension of logical network isolation across internetworks—typically through routing and Virtual Private Network (VPN) technologies operating at Layer 3 and above—is examined in Chapter 15.

14.3.3.3 Store-And-Forward Versus Cut-Through Switching

Switches may operate using different forwarding techniques:

Store-and-forward switching: The entire frame is received and checked for errors before forwarding.
Cut-through switching: Forwarding begins as soon as the destination address is read, reducing latency but potentially propagating corrupted frames.

Modern enterprise switches typically use store-and-forward operation to ensure error isolation.

14.3.3.4 Scalability And Hierarchical Design

Switching enables scalable hierarchical LAN architectures, typically organized into:

Access layer: Connects end devices.
Distribution layer: Aggregates access switches.
Core layer: Provides high-speed backbone interconnection.

In data centers, this hierarchy may be replaced or supplemented by spine–leaf architectures, which provide predictable low-latency paths and high bandwidth.

While switching determines how frames are forwarded logically within the LAN, achievable data rate and link distance are governed by the physical layer. We therefore next examine the evolution of Ethernet’s physical-layer signaling techniques.