Ethernet is a family of computer networking technologies standardised primarily through IEEE 802.3 and widely used for Local Area Network connectivity. It defines mechanisms for transmitting frames over various physical media and has evolved from early shared coaxial systems into modern switched, full-duplex networks using copper and optical fibre. Ethernet frames contain fields including destination and source MAC addresses, an EtherType or length field, payload and a frame check sequence. Ethernet can operate at many speeds, from historical low-speed implementations through Fast Ethernet, Gigabit Ethernet, 10 Gigabit Ethernet and much higher data rates. Despite substantial changes in physical implementation, the Ethernet family has retained a recognisable frame-based data-link architecture.
Early Ethernet, developed at Xerox PARC and subsequently standardised through the IEEE, used a shared-medium model based on Carrier Sense Multiple Access with Collision Detection. Devices listened before transmitting and detected collisions when multiple stations transmitted simultaneously. The widespread adoption of Ethernet switches fundamentally changed local Ethernet operation because each switched port generally forms an independent full-duplex link. Collision detection is therefore not used in the ordinary operation of modern full-duplex switched Ethernet. This historical transition is important because modern Ethernet should not be described as though all devices remain on one shared cable competing through CSMA/CD.
An Ethernet frame provides the fundamental data-link unit for ordinary Ethernet communication. The destination MAC address identifies the intended local recipient, while the source MAC address identifies the transmitting interface. The EtherType field can identify an upper-layer protocol such as IPv4 or IPv6, and the frame check sequence provides error detection. Ethernet itself does not provide end-to-end routing across arbitrary networks; it is normally used within a Layer 2 domain, while IP provides Layer 3 internetworking. Switches forward frames using MAC-address information, whereas routers make forwarding decisions based primarily on network-layer addresses and routing information.
Ethernet can use twisted-pair copper cabling, multimode fibre, single-mode fibre and other physical media depending on the standard. Copper variants such as 1000BASE-T and faster standards can use twisted-pair cabling with appropriate categories and installation requirements. Fibre variants provide longer reach and immunity to electromagnetic interference, making them particularly useful for building backbones, data centres and telecommunications environments. Ethernet physical standards specify details such as signalling, encoding or modulation, connector characteristics, reach and media requirements. A device advertised as supporting a particular Ethernet speed therefore still requires compatible cabling, transceivers and equipment at the other end of the link.
Ethernet switches learn MAC addresses by examining the source address of incoming frames and associating that address with the relevant switch port. If the destination address is known, the switch can forward the frame through the corresponding port; if it is unknown or the frame is a broadcast, the switch may flood it within the appropriate VLAN. This learning process is dynamic and can change when devices move or topology changes. Broadcasts and certain other frames remain relevant to the Layer 2 domain, which is one reason large flat Ethernet networks can become difficult to manage. VLANs provide logical separation by assigning frames to distinct broadcast domains within shared switching infrastructure.
Ethernet supports VLAN tagging through IEEE 802.1Q, allowing a frame to carry information identifying its VLAN when transported across a trunk link. This makes it possible for multiple logical networks to share the same physical links between switches. Ethernet also supports link aggregation through standards such as IEEE 802.1AX, allowing multiple physical links to operate as a logical aggregated connection. Power over Ethernet standards can deliver electrical power over suitable copper Ethernet connections, enabling devices such as access points, cameras and IP telephones to receive power and data through the same cabling. These extensions have significantly increased the flexibility of Ethernet beyond simple host-to-host data transfer.
Ethernet reliability and topology management depend on the design of the switching infrastructure and associated protocols. Redundant Layer 2 paths can create loops that cause frames to circulate indefinitely, so Spanning Tree Protocol and its variants can block selected paths and maintain loop-free forwarding. Modern routed architectures often reduce the size of Layer 2 domains and use Layer 3 routing between sections, while data-centre Ethernet fabrics can use equal-cost multipath routing and specialised designs to avoid large spanning-tree domains. Ethernet itself detects corrupted frames using the frame check sequence, but it does not normally retransmit corrupted frames at the Ethernet layer. Higher-layer protocols such as TCP may provide end-to-end recovery where required.
Ethernet became dominant because it offered an effective combination of interoperability, scalability, performance and comparatively economical infrastructure. Its standardisation allowed equipment from different manufacturers to communicate using common specifications, while successive generations increased speed without abandoning the fundamental Ethernet model. Ethernet now appears in office access networks, campus backbones, data centres, industrial environments, servers, storage systems and many other settings.
Its role has expanded from a technology associated mainly with small local networks to a broad family capable of very high-speed links over significant physical distances. The continued evolution of Ethernet demonstrates how a networking technology can retain a common architectural identity while changing substantially in physical signalling, speed, topology and operational context.
