Ethernet networking is a family of computer networking technologies used to transmit data between devices over wired network connections, primarily within local and data-centre networks. Ethernet is standardised mainly by the IEEE 802.3 family and operates principally at the physical and data-link layers of networking architectures. Ethernet defines mechanisms for framing data, addressing devices through MAC addresses, controlling access to the medium in applicable configurations and specifying physical transmission characteristics.
Modern Ethernet networks normally use switches and full-duplex links, with devices connected individually to switch ports through twisted-pair copper or fibre-optic cabling. Ethernet is not synonymous with the Internet, IP or a LAN, although it is one of the most important technologies used to build LANs that carry IP traffic. Ethernet can transport IPv4, IPv6 and many other protocols, and high-speed Ethernet is extensively used in data centres, enterprise networks, telecommunications infrastructure and Internet exchange environments.
Ethernet originated in the 1970s as a local networking technology developed at Xerox PARC, with early work associated with Robert Metcalfe and colleagues. The original Ethernet concept used a shared coaxial medium and a carrier-sense multiple-access with collision detection mechanism, commonly abbreviated CSMA/CD. Devices sharing the medium listened before transmitting and detected collisions when multiple devices transmitted simultaneously.
Ethernet evolved through standardisation and commercial adoption, eventually moving from shared coaxial networks to twisted-pair star topologies using hubs and then to switched Ethernet. This transition fundamentally changed how Ethernet operated. Modern switched full-duplex Ethernet links generally do not experience the collisions that characterised early shared-medium Ethernet, so CSMA/CD is largely a historical mechanism in contemporary switched networks. The term Ethernet nevertheless covers a very broad family of standards developed over several decades.
An Ethernet frame is the basic data-link-layer unit used to carry information across an Ethernet network. A traditional Ethernet frame contains destination and source MAC addresses, an EtherType or related length/type field, a payload and a frame check sequence used for error detection. The MAC address is a link-layer identifier associated with a network interface, although modern virtualisation and network configurations can introduce additional complexities. The EtherType can identify the protocol carried within the payload, such as IPv4 or IPv6.
Ethernet's frame check sequence allows the receiver to detect corruption of the frame during transmission, although Ethernet itself does not generally retransmit corrupted frames at the link layer in the way TCP retransmits lost data. Higher-layer protocols can provide reliability when required. Ethernet frames are therefore distinct from IP packets and TCP segments: the frame belongs to the data-link layer, while the IP packet and TCP segment operate at higher layers.
Modern Ethernet relies heavily on switches. A switch examines the destination MAC address of an incoming frame and forwards it through the appropriate port based on its MAC-address forwarding database. The switch learns source MAC addresses by observing incoming frames and associating those addresses with ports. If the destination is unknown or broadcast traffic is involved, the switch may forward the frame across multiple relevant ports according to Ethernet's forwarding rules.
Each switched port normally represents a separate collision domain, and full-duplex operation allows simultaneous transmission and reception. Managed switches can provide VLANs, link aggregation, quality-of-service controls, port security, monitoring and other functions. Layer 3 switches can additionally perform IP routing, allowing a single device to combine high-speed switching and routing functions.
Ethernet supports a very wide range of speeds and physical media. Historical Ethernet operated at 10 megabits per second, followed by Fast Ethernet at 100 megabits per second and Gigabit Ethernet at 1 gigabit per second. Modern standards include 2.5 and 5 gigabit Ethernet for certain twisted-pair environments, 10 gigabit Ethernet and much higher-speed technologies such as 25, 40, 50, 100, 200 and 400 gigabit Ethernet, with standards continuing to evolve for specialised environments.
Copper twisted-pair Ethernet commonly uses RJ45-style connectors in familiar enterprise and residential deployments, while fibre-optic Ethernet is widely used for longer distances, high-speed uplinks and data-centre interconnections. The actual distance supported depends on the physical-layer standard, cable type, transceiver characteristics and installation conditions. Ethernet therefore describes a technology family rather than one fixed speed or cable type.
Ethernet's relationship with IP is important because the two technologies solve different problems. IP provides logical addressing and routing between networks, whereas Ethernet provides local data-link connectivity and framing. A computer can send an IP packet to another device on the same subnet by encapsulating it in an Ethernet frame and addressing the frame toward the appropriate MAC address.
If the destination is on another IP network, the Ethernet frame is normally addressed to the local router's MAC address, and the router forwards the IP packet into another link-layer network. Address-resolution mechanisms determine the relevant local link-layer address. This layered design allows IP to operate over Ethernet without making IP dependent on Ethernet everywhere. IP can also operate over Wi-Fi, fibre-based technologies and other link layers. Ethernet is consequently one component of a larger networking architecture rather than the complete networking stack.
Ethernet offers several practical advantages, including predictable wired connectivity, high capacity, low latency, low packet loss under appropriate conditions and resistance to many forms of radio-frequency interference that affect wireless networks. A wired Ethernet link does not require shared radio spectrum, so nearby Wi-Fi networks do not directly consume its physical transmission medium. Full-duplex switched Ethernet also allows simultaneous bidirectional traffic without the contention mechanisms associated with shared wireless media.
However, Ethernet requires physical cabling or fibre infrastructure, which can increase installation cost and reduce mobility. Cable length, physical damage, connector quality, electromagnetic environment for copper and transceiver compatibility can all affect reliability. Fibre provides strong immunity to electromagnetic interference and can support long distances, but it generally requires different transceivers and installation practices from copper.
Ethernet networking remains a foundational technology because it combines standardised framing, addressing and physical transmission with an extensive ecosystem of switches, network interfaces, cables, optical transceivers and management tools. It is used in homes for wired devices, in businesses for access and backbone connectivity, in data centres for server and storage networks, and in service-provider infrastructure for high-capacity transport and aggregation.
VLANs allow multiple logical Layer 2 networks to share physical Ethernet infrastructure, while routing connects those networks at Layer 3. Ethernet has therefore evolved far beyond its original shared coaxial implementation: modern Ethernet is predominantly switched, full-duplex and highly scalable, with speeds ranging from low-gigabit residential connections to hundreds of gigabits per second in specialised infrastructure. Understanding Ethernet requires distinguishing frames from IP packets, MAC addresses from IP addresses, switches from routers, physical media from logical networks, and historical shared-medium Ethernet from contemporary switched Ethernet.
