Ethernet has evolved to include higher bandwidth, improved
medium access control methods, and different physical media. The
multidrop coaxial cable was replaced with physical point-to-point links connected by
Ethernet repeaters or
switches. Ethernet stations communicate by sending each other
data packets: blocks of data individually sent and delivered. As with other IEEE 802 LANs, each adapter comes programmed with a globally unique 48-bit
MAC address so that each Ethernet station has a unique address. The MAC addresses are used to specify both the destination and the source of each data packet. Ethernet establishes link-level connections, which can be defined using both the destination and source addresses. On reception of a transmission, the receiver uses the destination address to determine whether the transmission is relevant to the station or should be ignored. A network interface normally does not accept packets addressed to other Ethernet stations. An EtherType field in each frame is used by the operating system on the receiving station to select the appropriate protocol module (e.g., an
Internet Protocol version such as
IPv4).
Ethernet frames are said to be
self-identifying, because of the EtherType field. Self-identifying frames make it possible to intermix multiple protocols on the same physical network and allow a single computer to use multiple protocols together. Despite the evolution of Ethernet technology, all generations of Ethernet (excluding early experimental versions) use the same frame formats. Mixed-speed networks can be built using Ethernet switches and repeaters supporting the desired Ethernet variants. Due to the ubiquity of Ethernet and the ever-decreasing cost of the hardware needed to support it, by 2004 most manufacturers built Ethernet interfaces directly into
PC motherboards, eliminating the need for a separate network card.
Shared medium adapter, a similar model transceiver with a
10BASE5 adapter, an
AUI cable, a different style of transceiver with 10BASE2
BNC T-connector, two 10BASE5 end fittings (
N connectors), an orange
vampire tap installation tool (which includes a specialized drill bit at one end and a socket wrench at the other), and an early model 10BASE5 transceiver (h4000) manufactured by DEC. The short length of yellow 10BASE5 cable has one end fitted with an N connector and the other end prepared to have an N connector shell installed; the half-black, half-grey rectangular object through which the cable passes is an installed vampire tap. Ethernet was originally based on the idea of computers communicating over a shared coaxial cable acting as a broadcast transmission medium. The method used was similar to those used in radio systems, with the common cable providing the communication channel likened to the
Luminiferous aether in 19th-century physics, and it was from this reference that the name
Ethernet was derived. The original Ethernet's shared coaxial cable (the shared medium) traversed a building or campus to connect every attached machine. A scheme known as
carrier-sense multiple access with collision detection (CSMA/CD) governed the way the computers shared the channel. This scheme was simpler than competing Token Ring or
Token Bus technologies. Computers are connected to an
Attachment Unit Interface (AUI)
transceiver, which is in turn connected to the cable (with
thin Ethernet, the transceiver is usually integrated into the network adapter). While a simple passive wire is highly reliable for small networks, it is not reliable for large extended networks, where damage to the wire in a single place, or a single bad connector, can make the whole Ethernet segment unusable. Through the first half of the 1980s, Ethernet's
10BASE5 implementation utilised a coaxial cable in diameter, later referred to as
thick Ethernet or
thicknet. Its successor,
10BASE2, called
thin Ethernet or
thinnet, used the
RG-58 coaxial cable. The emphasis was on making installation of the cable easier and less costly. Since all communication happens on the same wire, any information sent by one computer is received by all, even if that information is intended for just one destination. The network interface card interrupts the
CPU only when applicable packets are received: the card ignores information not addressed to it. Use of a single cable also means that the data bandwidth is shared, such that, for example, available data bandwidth to each device is halved when two stations are simultaneously active. A collision happens when two stations attempt to transmit at the same time. They corrupt transmitted data and require stations to re-transmit. The loss of data and retransmission reduce throughput. In the worst case, where multiple active hosts connected with maximum allowed cable length attempt to transmit many short frames, excessive collisions can reduce throughput dramatically. However, a
Xerox report in 1980, published in
Communications of the ACM, studied the performance of an existing Ethernet installation under both normal and artificially generated heavy load. The report claimed that 98% throughput on the LAN was observed. Somewhat larger networks can be built by using an
Ethernet repeater. Early repeaters had only two ports, allowing, at most, a doubling of network size. Once repeaters with more than two ports became available, it was possible to wire the network in a
star topology. Early experiments with star topologies (called
Fibernet) using
optical fiber were published by 1978. Shared cable Ethernet is always hard to install in offices because its bus topology is in conflict with the star topology cable plans designed into buildings for telephony. Modifying Ethernet to conform to twisted-pair telephone wiring already installed in commercial buildings provided another opportunity to lower costs, expand the installed base, and leverage building design, and, thus, twisted-pair Ethernet was the next logical development in the mid-1980s. Ethernet on unshielded
twisted-pair cables (UTP) began with
StarLAN at in the mid-1980s. These evolved into 10BASE-T, which was designed for point-to-point links only, and all termination was built into the device. This changed repeaters from a specialist device used at the center of large networks to a device that every twisted pair-based network with more than two machines had to use. The tree structure that resulted from this made Ethernet networks easier to maintain by preventing most faults with one peer or its associated cable from affecting other devices on the network. Despite the physical star topology and the presence of separate transmit and receive channels in the twisted pair and fiber media, repeater-based Ethernet networks still use half-duplex and CSMA/CD, with only minimal activity by the repeater, primarily the generation of the
jam signal in dealing with packet collisions. Every packet is sent to every other port on the repeater, so bandwidth and security problems are not addressed. The total throughput of the repeater is limited to that of a single link, and all links must operate at the same speed. This reduces the forwarding latency. One drawback of this method is that it does not readily allow a mixture of different link speeds. Another is that packets that have been corrupted are still propagated through the network. The eventual remedy for this was a return to the original
store and forward approach of bridging, where the packet is read into a buffer on the switch in its entirety, its
frame check sequence verified and only then the packet is forwarded. When a twisted pair or fiber link segment is used, and neither end is connected to a repeater,
full-duplex Ethernet becomes possible over that segment. In full-duplex mode, both devices can transmit and receive to and from each other at the same time, and there is no collision domain. This doubles the aggregate bandwidth of the link and is sometimes advertised as double the link speed (for example, for Fast Ethernet). The elimination of the collision domain for these connections also means that all the link's bandwidth can be used by the two devices on that segment and that segment length is not limited by the constraints of collision detection. Since packets are typically delivered only to the port they are intended for, traffic on a switched Ethernet is less public than on shared-medium Ethernet. Despite this, switched Ethernet should still be regarded as an insecure network technology, because it is easy to subvert switched Ethernet systems by means such as
ARP spoofing and
MAC flooding. The bandwidth advantages, the improved isolation of devices from each other, the ability to easily mix different speeds of devices and the elimination of the chaining limits inherent in non-switched Ethernet have made switched Ethernet the dominant network technology.
Advanced networking Simple switched Ethernet networks, while a great improvement over repeater-based Ethernet, suffer from single points of failure, attacks that trick switches or hosts into sending data to a machine even if it is not intended for it, scalability and security issues with regard to
switching loops,
broadcast radiation, and
multicast traffic. Advanced networking features in switches use
Shortest Path Bridging (SPB) or the
Spanning Tree Protocol (STP) to maintain a loop-free, meshed network, allowing physical loops for redundancy (STP) or load-balancing (SPB). Shortest Path Bridging includes the use of the
link-state routing protocol IS-IS to allow larger networks with shortest path routes between devices. Advanced networking features also ensure port security, provide protection features such as MAC lockdown and broadcast radiation filtering, use
VLANs to keep different classes of users separate while using the same physical infrastructure, and use
link aggregation to add bandwidth to overloaded links and to provide some redundancy. In 2016, Ethernet replaced
InfiniBand as the most popular system interconnect of
TOP500 supercomputers. In many industrial systems, Ethernet and
fieldbus coexist, each performing certain roles, and
data is exchanged between them through
gateways. ==Varieties==