Ethernet Hardware and Topologies

 

http://pclt.cis.yale.edu/pclt/COMM/ETHER.HTM- old but very clear on the basics

http://www.ethermanage.com/ethernet/ethernet.html– Charles Spurgeon's ethernet site

 

Ethernet is set of hardware and firmware standards defining a LAN (Local Area Network) networking protocol. This protocol was originally half-duplex and packet switched and transmitted packets via CSMA/CD. These characteristics of Ethernet will be discussed later. About 85% of networks are Ethernets, and it is the type supported by default by the adapter firmware and hardware on Sun computers. Ethernet is also cheap. It costs about $50 a station to hook one up, as opposed to $1000 for a host with ATM (Asynchronous Transfer Mode)..

 

Ethernet was invented at by Dr. Robert Metcalfe at "Xerox PARC", the Xerox Palo Alto Research Center, which was one of the most important private institutions in the development of the Internet. The first specifications for the Ethernet were called "DIX" Ethernet after the corporations (DEC, Intel, Xerox) that cooperated in their creation. In 1985 the IEEE took over the definition of Ethernet standards with the publication of the IEEE 802.3 "Ethernet" standards. Ethernet includes hardware standards  (wiring, interface), interface firmware standards, and Ethernet packet (called "frame") standards. We will look at the hardware standards for Ethernet in this document, then briefly consider other network types. In the next document we will discuss firmware and packet standards.

 

LANs: A LAN, or Local Area Network, is defined as a network restricted to one campus or other small area.  It cannot extend over more than 2000 m, and does not transmit over cables that are shared with other than its own users.  In ordinary use, a LAN is a network where all hosts are connected and can communicate with each other entirely over hardware, without going through a router. A router is a device that uses software or firmware (but not hardware) to forward packets from one LAN to another.

 

In contrast, a WAN, or Wide Area Network, is defined as a network formed of LANs connected through routers.  WANs transmit over lines that are leased and shared, and can span the globe. LANS allow sharing of resources, centralization of management and permits workgroups to have their own virtual workspace.  A VLAN (Virtual Local Area Network) is a local area network that runs on the same switch as another local area network. Setting up a VLAN requires a switch that supports this (VLAN-enabled), since one subset of the ports on the switch are assigned to one network, and the others to a different network. These separate traffic and thus improve security. VLANs also increase flexibility - you can move a computer from one network to another without changing any hardware.

 

LAN Topologies

 

Backbones and hubs: The original Ethernet standard was half-duplex. Half-duplex networks immediately disperse all packets to all hosts on the network over a single hardware device. As a result, a packet cannot share the wire with any other packet, and hosts on the network can send or receive packets, but never do both. The original Ethernet carried signals on coaxial cable, which has a single copper core wire, and transmitted data at 10 Mbps (Megabits per second).  Full-duplex Ethernet is now widely implemented. Hosts using full-duplex Ethernet can send and receive packets at the same time.

 

Half-duplex Ethernet assumes that all the hosts on a network are somehow connected to a central communication point.  This implies that they are connected by cable (or using a wireless receiver) to a device that allows them to hear all the transmissions on the network.  In networks set up in the late 1990s and later, this central point is generally a hub.  The hub is a half-duplex, multi-port device to which hosts on a network connect. It immediately repeats all incoming packets down all attached cables, including the cable attached to the originating host. Networks connected to a hub (or switch) are said to have a "star" topology, where topology simply refers to the arrangement of systems and connectors in a network.  In the 1980s and early 1990s, the central point was generally a backbone.  The backbone was a thick coaxial cable similar to TV cable. Cables from each host tapped into the backbone directly.  All transmissions were heard on all cables attached as well as the backbone. Networks using a backbone are said to have "bus" topology. 

 

Non-Ethernet networks:  An Ethernet topology is either "bus" – a backbone – or "star" – a hub. Other types of networks exist, and they may use other topologies.  In a ring, each computer is connected to the next. Packets sent from one are picked up by the next. If the next computer is not the destination of the packet, it passes the packet to the next computer in the series.  Ring topologies are generally "token" passing. An empty packet, called a "token" is passed from host to host. When a host has a packet to send, it waits til it has the token, then sends out its packet. 

 

LAN Hardware

 

Cable naming: LAN cable standards are given names that reflect the physical properties of the cabling.  In general, the name is something like XBASE-Y, where X is 10, 100 or 1000, and represents the speed of transmission on the cable in Mbps, BASE means that the cabling is baseband and not broadband, while Y indicates the type of cabling. Cable standards include

 

10BASE-2 – coax that can stretch 2 x 100 meters (thin coax)

10BASE-5 – coax that can stretch 5 x 100 meters (thick coax)

T, as in 10BASE-T, 100BASE-TX – twisted pair cable

F, as in 100BASE-FX – fiber optic cable

Other letters or numbers are used as the manufacturer decides.

 

Types of cabling: Coaxial or "coax" cable was the earliest networking cable defined, and is still used for backbones and in some limited settings in networks.  It can carry signals for long distances, and thus may be used to connect subnets, but it is always half-duplex and cannot exceed transmission speeds of 10 Mbps.  Coax cable is still used by cable modem and cable TV, although those cables do not meet the impedance standards for Ethernet and the cables are not interchangeable.  Coax has four layers: a copper or copper clad core, surrounded by a layer of insulation, then by a braided metal shield and a finally by a PVC or Teflon jacket. The transmissions flowing through the copper core are therefore highly shielded.  It is also fairly expensive. 

 

Twisted pair cable is much cheaper than coaxial cable, and it can be used in a great variety of networks.  For example, the same kind of twisted pair wire can be used to set up 100 Mbps Ethernet or ATM networks. (The type of cable has little to do with the network specifications; it is the network interface that dominantly determines how the network functions.) In twisted pair cabling, a pair of wires is twisted together for 2-3 twists per foot (category 3) or 2-3 twists per inch (for category 5). The twists decrease EM interference, or "crosstalk". Category 5 cable will carry 10 or 100 mbps signals.  Twisted pair cable can carry a signal for about 100 meters, and connects to an RJ-45 plug on each end, which in turn goes to an internal transceiver or a hub. The original 10BASE-T specification included 2 pairs of wires: one pair for transmission, and one pair for receiving. 

 

CAT-5e cable is the same as CAT-5 cable, but it is manufactured to a higher standard so it can be used for gigabit ethernet (1000 mbps).. CAT 6 cable also has four pairs of wires and is limited to distances of less than 100 meters , but it is manufactured to higher standards still.  It is also used for gigabit ethernet because it loses far fewer packets due to reduced crosstalk, (about a factor of 10) and has double the bandwidth. Crosstalk is noise on a wire caused by transmissions on adjacent wires. The speed of communication of CAT 6 cable is double that of  CAT 5e cable.  It is more expensive than CAT 5 cable, and applications are generally not yet able to take advantage of the higher speed, so it isn't very widely used.

 

Twisted pair cable can be used for a 100Mbps connection with 2 pairs of wires, allowing for half duplex, or 4 pairs (100BASE-T4) allowing for full duplex. Full duplex also requires a full-duplex capable interface, and an application that can make use of it.  The 1000BASE-T standard also exists.  10BASE-T and 100BASE-TX both run over CAT5 twisted pair wire.

 

More than you wanted to know: How are 10 Mbps and 100 Mbps different electrically? 10 uses simple equalization. 100 is more complicated. Only the ends of the wire are involved, so it doesn't matter what kind of wire it is. The interface does the equalization. Electrical signals get dispersed (according to wavelength) as they travel down a cable.  This has to be removed on the receiving end. Removing it slowly is much easier than removing it quickly!

 

Fiber optic cabling: Only one standard of 10BASE fiber optic cabling is widely used: FL, which is used in Internet backbones. Fiber optic cable transmits data as bursts of light produced by lasers rather than as electrical impulses. You can use fiber optic cable in an Ethernet LAN, but it works best for connecting LANs.  There are also 100 Mbps and 1000 Mbps standards for fiber optics, for both short and long wavelength laser. Original fiber optic cable used glass and lacked flexibility, and it was expensive, so it wasn't practical for anything but backbones. Polymer fiber is now available and is very flexible.  It has also come down in price.

 

Crossover cables: Transmission wires must go to the receiving wires on the hub or switch. Usually a hub will be set up to send on some pins of the cable, and the hub is set up to receive on those pins. As a result the wires on most cables go straight through. The name of such a cable is also "straight-through." If a cable connects two hosts or two hubs, the same pins on each host (or hub) will send and to receive, so a straight-through cable will connect transmitting pins to transmitting pins and receiving pins to receiving pins, and will not work. Connecting two hosts or hubs requires a cable where transmission and reception wires are crossed within the cable. Such a cable is a "cross-over" cable.  Some hubs can be switched back and forth depending on the kind of cable used. Check the two ends of a twisted pair cable to see if the wire colors are laid out the same (straight-through, usually with a grey coat) or have been changed around (crossover, usually orange) on either end of the cable.

 

Ethernet interfaces: One end of each cable must connect to a port on the computer. These ports are generally on NICs (network interface cards) or are built into the motherboard. They may be called adaptors, interfaces, or network controllers.  Sun computers come with an interface on the motherboard, and more can be added on NIC cards. The interface is crucial in determining what kind of transmission can occur on the network.  Twisted pair wires will work in full or half duplex transmissions, at 10, 100 or 1000 Mbps speeds, but interfaces will not.  Sun's interfaces, starting with the hme series, are full-duplex capable. 

 

Network devices: The other end of a cable is may be connected to a hub, which is a passive, multi-port, half-duplex device that repeats all incoming packets out all ports. A cable can also connect into a bridge, which connects to another cable on the other side. Bridges are most useful in extending the length of backbone cable.  A bridge filters traffic, passing through only packets destined for the other side of the bridge.  It also discards bad packets, so it is smarter than a hub, which is a completely passive device. A length of cable can also be extended with a repeater, which is a dumb device like a hub that regenerates a signal and passes it on the other side. 

 

In modern LANs, switches (or "switching hubs") are more commonly used than hubs.  Like a bridge, a switch filters traffic, except for multiple hubs or hosts. A switch analyzes a packet, determines the destination and repeats the packet only through the port attached to the destination host. A VLAN (Virtual Local Area Network) switch can be configured so that selected ports are assigned to separate networks. It is therefore possible to set up multiple networks on a single switch, which is cheaper, more convenient, and as secure as a separate LAN. A system on a VLAN can be moved to another network without making any hardware changes - the administrator simply reconfigures the switch so that the system's port is on the new network. A router may be attached to two different ports on the same switch and route between the two networks, since the switch may not be a router.

 

Acronyms:

AA – ATM Adaptation

ATM – Asynchronous Transfer Mode - a type of backbone topology in which 53 byte cells are passed. It has a backbone of cables and switches and uses adaptor cards.  It is used to create a virtual channel with guaranteed quality of service, generally for teleconferencing.  It has a dedicated path and bandwith and a continuous connection. 

CRC – Cyclical Redundancy Check – checksum for a data packet.

CSMA/CD – Carrier Sense Multiple Access/Collision Detect – a transmission method for a network – used by Ethernet. Multiple hosts are connected to a single cable or set of cables (Multiple Access). Hosts check the wire and transmit when it is not busy (Carrier Sense).  Packets are broadcast, and the originating system then listens for the packets it sent out.  If it doesn't hear its own packets, it waits for a random time period, then rebroadcasts (Collision Detect).

EIA - Electronic Industries Association - standards organization

FDDI - Fiber Distributed Data Interface – a standard for fiber optic cable, which is made of fibers of light-transmitting material such as glass or polymer.

FOIRL – Fiber Optic Inter Repeater Link – original specification for fiber optic links between Ethernet segments.

ISO  - International Organization for Standardization - standards organization – the umbrella organization for ITU and ANSI. ISO developed the ISO/OSI model.

LAN – Local Area Network

LANE – Local Area Network Emulation

LED – Light Emitting Diode.

MDI – Medium Dependent Interface – the connector such as RJ45 or vampire clamp.

MDI-x – connectors on crossover cable.

MII – Media Independent Interface - a set of pins that sits next to the normal twisted pair input on several Sun ethernet interfaces (both built in to mother boards and to IO boards and regular cards) These can plug into the thick and thin wire coaxial cable connectors. They require a "media-dependent" cable to attach to the transceiver.

nf – new fiber

            single – connects to one ring – half duplex

            double – connects to 2 rings – full duplex

NIC – Network Interface Card – card with ethernet ports

PDU – Packet Data Unit

QOS – Quality of Service

STP – Spanning Tree Protocol

TIA – Telecommunications Industry Association – a hardware standards organization.

UTP – Unshielded Twisted Pair – a type of cable.

VLAN - Virtual Local Area Network - a local area network that runs on the same switch as another local area network. Setting up a VLAN requires a switch that supports this (VLAN-enabled), since one subset of the ports on the switch are assigned to one network, and the others to a different network. These separate traffic and thus improve security.

 

Definitions:

asynchronous traffic – traffic which goes in bursts. FDDI supports this as well as synchronous traffic, but ATM does not support async traffic.

autonegotiation – sending system runs the fast link pulse test, sending a 16 bit word at a given speed.  If there is no reply it tries sending the same word at a lower speed until it gets a response from its hub. 

backbone – a network topology where a single coaxial cable carries transmissions from hosts connected to it by individual cables.  This topology was typical about 10-15 years ago, but is no longer commonly set up as part of a LAN. Backbones are still used in the Internet.

baseband – a single frequency.  Signal is sent as a set of 1s and 0s in whatever frequencies are generated by the source.

bridge – a device which connects two segments of backbone.  It uses the gratuitous ARP to collect information from the hosts about which side of the bridge each is on.  It then transmits only packets meant for the other side of the bridge. All others propagate on the cable only on the side of the bridge on which they originated. It thus isolates traffic into 2 collision domains.  A bridge checks packets, discarding bad ones before it retransmits.

broadband – cable carrying a signal which is modulated to another high frequency. This reduces the need for equalization.  Fibers whose diameter is many times greater than the wavelength of light transmitted are broadband.  The term is also used to indicate cable which  allows 1 Mbit or greater bandwidth such that different services can use it at the same time by utilizing different frequencies.

category 3 twisted pair – 2-3 twists/foot voice grade cable

category 5 twisted pair – 3-5 twists per inch data grade cable.

concentrator – a device which serves the tasks of multiple devices, such as a repeater and a hub, or a switch, a router and a gateway.

crossover cable – a transmit pair is crossed to a receive pair: 1,3 to 2,6.

ethernet – a high speed, inexpensive LAN technology, consisting of cables, hubs, repeaters, bridges, switches and routers.  Standards are set by IEEE.

full duplex – it is possible to send and receive over the cable at the same time, because the cable has 4 or more pairs of wire, the interface is capable of decoding such communications, and only one system transmits at a time. In Ethernet, full duplex is possible only with point-to-point connections or between a workstation and a switch port, such that an isolated collision domain is formed between only 2 devices.

half duplex – it is possible to send and receive over the cable, but at different times. All CSMA/CD transmissions must be half duplex.

hub – a multiport bridge which connects systems in a star topology. By definition they are half-duplex.

gateway – a system which translates data from one protocol to another.

gratuitous ARP – an ARP sent out by a system at boot time with its MAC address.

LAN – Local Area Network – a network owned by a single entity, 2000 meter maximum extent, does not go over leased, shared lines.  LANs use the CSMA/CD transmission method.

multimode – a type of fiber optic transmission that uses an LED to generate light. 

packet-switched – a network in which hosts take turns transmitting packets.

put-through switching: packets are transmitted without any filtering.

point-to-point connection – a connection made directly between two devices.

repeater – a device which connects two sections of cable and regenerates an incoming signal, thus extending the length of a network.  No filtering occurs.

router – a system or device connected to more than one network which reads IP addressing information and routes packets to the correct network.

segment – a section of cable extending between MDIs.

single mode – a type of fiber optic transmission that uses a laser to generate light. Fibers only a few times larger in diameter than the wavelength of light transmitted is single mode.

straight-through cable – cable that has no crossing of wires.

straight-through switching - the switch or bridge reads the CRC on the packet to see if the packet is ok before it transmits it.

switch – a multiport bridge which isolates traffic into multiple collision domains. Since it can have multiple paths going at once, it is fast.

synchronous traffic – traffic which flows constantly on a dedicated channel. FDDI supports this, as does ATM.

token passing – access method on a ring topology. Systems pass a token one to another. The system possessing the token may transmit. All others are silent.

topology – the organization and connection design of a network.

WAN – any size, goes over leased lines and connects LANs

 

 

Misc:

 

Hardware standards:

 

standard

material

duplex

IEEE std #

max length (m)

notes

1000BASE-LX

fiber

full

 

300-550

long wavelength

1000BASE-SX

fiber

full

 

300-550

short wavelength

1000BASE-T

twisted pair (cat. 5e)

full

802.3ab

550-3000

 

1000BASE-CX

heavy copper

full

 

100

 

1000BASE-X

fiber, twisted pair

full/half

802.3z

25

 

100BASE-FX

fiber

full

 

100

 

100BASE-T4

twisted pair

full

 

100

4 pairs of wires

100BASE-TX

twisted pair (cat. 5)

half

802.3u

100

2 pairs of wires- "fast"ethernet

10BASE-FL

fiber

full

 

2000

fiber backbone.

10BASE-T

twisted pair

 

802.3

100

2 pairs of wires

internal transceiver to RJ45 jack to hub.

10BASE-2

thin coaxial cable (about 6 mm)

half

 

200

 

10BASE-5

thick coaxial cable

(about 12 mm)

half

 

500

then: vampire clamp to backbone

now: used for medium length, highly reliable backbone connections.

 

Hardware differences:

hub vs repeater:  a hub is multi-port, a repeater is two port.

bridge vs switch: a bridge has two ports, so isolates network on either side into 2 collision domains. A switch is multi port, and so isolates the network into multiple collision domains.  Both filter, store packets and extend the link-local area. Max. bridges before a router: 7.

Network topologies:

star – many segments lead into one hub.

bus – a single cable has many segments attached to it.

ring – (these are not Ethernet) – systems are connected in a ring and pass around the ring

Hosted by www.Geocities.ws

1