Ethernet and Cabling
- What is Ethernet?
- Sharing the Network - CSMA/CD
- Manchester Encoding
- Ethernet variations
- 10Base2
- 10/100BaseT
What is Ethernet?
As with Token Ring described earlier, Ethernet describes another standard local area networking implementation. Originally invented by Xerox in the early 70s, today it is by far the cheapest and most common approach.
Sharing the Network - CSMA/CD
Like token ring networks, there must be a mechanism to eliminate noise on the network, yet ensure that a message reaches its intended destination. There must be a way to coordinate messages so that all nodes get a fair shot at communicating over the shared cable.
In Ethernet networks, there are no tokens and any node can attempt to send data at any given time. This would result in collisions, except that Ethernet employs a media access method called CSMA/CD (Carrier Sense Multiple Access with Collision Detect) to avoid them. Basically, all nodes listen to the cable to see if any data is being transmitted. If a node detects data being transmitted on the LAN, it will not send its own data at that time. If there is no traffic on the LAN, it will send its data. In short, only one node is allowed to transmit at any given time; all others must wait for 'silence'.
How does a node know if another node is transmitting? Simple: if a node is transmitting, then the message being transmitted will be in the form of an electrical signal. The electrical signal is known as a 'carrier'. If no message is being transmitted, then there is no carrier on the wire. Thus, the process of checking if a signal is present on the wire is called carrier sense.
(By the way, a single 'lump' of data sent by a node across the LAN is called a frame, or more generically: a packet.)
But what happens if two nodes each have a message to send, and both check for a carrier and find the cable idle? These two (or more) nodes - upon ascertaining that there are no other network transmissions - attempt to send their frames at the same time. The result is a collision. When a collision occurs, the data transmitted by the two (or more nodes) interferes and results in a garbled message. Each transmitting node can detect a collision by comparing the message on the wire to the message it just transmitted. If there's a difference, chances are it was due to a collision.
When such a collision occurs, a jamming signal is sent out which delays transmissions from all nodes for a random period of time (determined using a 'backoff' algorithm). After the backoff period, the nodes retransmit their frames.
Why is the backoff period random? Well, if it was not, the transmitting nodes would all wait the same period of time and then retransmit at the same time... Resulting in another collision. By using a random backoff time, it is most likely that the nodes will wait for different periods. In short, this randomising approach significantly reduces the probability of two or more nodes attempting to send data across the network at the same time.
Of course, using random backoff times can not guarantee that a second collision won't occur. On a highly congested network, multiple collisions may occur. However, with each subsequent collision, the maximum amount of time allowed for the backoff period is doubled. This is called binary exponential backoff. Thus, there is more scope for the random times to be different between nodes.
Manchester Encoding
Ethernet uses Manchester encoding to transmit binary data. In a nutshell, this means that the binary value 1 is signalled by a rise in voltage, while a 0 is encoded by a voltage drop.
Recall that Ethernet networks send data in packets called frames. Ethernet frames must have a defined beginning. Manchester encoding defines the use of a pre-amble - a fixed 64 bit sequence which indicates where an Ethernet frame begins. For more details, check out the page on Ethernet frame structure in the TCP/IP section.
Ethernet variations
As if networking wasn't confusing enough...
There are several different variations of Ethernet, which primarily refer to the type of cabling used. Broadly speaking, these days there are really only two types of cabled network we need to concern ourselves with: the ageing 10Base2 (aka Thin Ethernet or Cheapernet) and the ubiquitous 10/100BaseT (sometimes known as Twisted-Pair Ethernet). In both cases, the initial 10 or 100 refers to the maximum data transfer rate of 10 or 100 megabits per second (i.e. millions of bits per second), respectively.
The word "Base" refers to the signalling transmission method of Ethernet, which is baseband. Baseband is a digital signal that is cheap to implement. This differs from broadband which is a modulated analogue signal and has the drawback of requiring a modem at each end of the line to convert the signal back to digital. (This is why cable modem broadband users need a cable modem: the broadband signal is converted by the cable modem into a digital signal that can be interpreted by an Ethernet card.)
Unfortunately, the character following the word Base has a less consistent meaning. First, let's look at 10Base2 Ethernet.
10Base2
10Base2 uses thin coaxial cable and in this case the 2 refers to the maximum supported coaxial cable length in the network in hundreds of metres. Hence 10Base2 supports upto 200m of cable. (In fact, this is a round-up: the maximum length is 185m in 10Base2.) This limitation on cable length is a physical constraint resulting from the fact that electrical signals become weaker with distance, as a result of resistance in the copper wire and EM radiation.
Similarly, the more robust 10Base5 Ethernet, which uses thick coaxial cabling, supports up to 500m of cable. However, this type of Ethernet will not be discussed any further.
The topology (a term used to describe the arrangement of nodes within a network) for a 10Base2 network is what is called the bus arrangement:
The bus topology of 10Base2 Ethernet networks
Each node is attached to the next by a single length of coaxial cable. At the nodes at each end, the cable must be terminated using a 50 Ohm BNC terminator. The terminator prevents electrical signals from 'reflecting' at the end of the cable. Reflections cause interference which would be perceived by sending nodes as collisions. Thus the network will not operately correctly without the terminators. These metallic connectors use a push-screw mechanism for attaching to the network card (aka network interface card or NIC - more on these later).
A BNC terminator used in 10Base2 Ethernet networks
All other nodes require a T-piece attached to the network card. This allows the cable from a node on one side to go "in", and a cable to the node on the other side to go "out". Hence if we look at the bus diagram above, we can see that each 'T-junction' where a node joins the main 'bus' represents a T-piece attached to a network card.
A T-Piece used to interconnect nodes in a 10Base2 Ethernet network
The major drawback of this bus arrangement (other than its antiquated status) is that any break in the line will cause the entire network to fail. Therefore all lengths of cable, nodes and network cards must be operational (at least as far as the network is concerned).
10/100BaseT
In 10/100BaseT, the T actually refers to the type of cabling used, which is telephone-type unshielded twisted-pair cabling.
Unshielded Twisted-Pair cabling
10/100BaseT networks support a cable length of about 100m. However, in this case we are not referring to the maximum total length of cable, but rather the distance between any node and its hub or switch. These devices will be discussed in more detail later, but for now it is sufficient to think of a hub as the device that provides the connectivity at the centre of a star topology network.
You see, unlike the bus arrangement used in 10Base2, 10/100BaseT Ethernets utilise the star arrangement:
The Star topology used in 10/100BaseT Ethernet networks
The hub is simply a device for interconnecting several nodes. Ethernet cabling from each node plugs into a port in the hub. The hub works by forwarding data from one node (entering the hub at a particular port) to all other nodes attached to it (by forwarding the data to all of its ports). This arrangement is preferable to the bus arrangement, because the loss of one node will not affect any other nodes attached to the hub. The only single point of failure that will kill the entire LAN is the hub itself, and these are cheap to replace.
This type of network is easily recognised from the fact that each length of cable ends with a RJ-45 modular connector (which strongly resemble the smaller RJ-11 telephone connectors).
RJ-45 connector
What's next
Now let's look at some network hardware, such as network cards, modems, hubs, switches and routers.
| Just Too Good Last updated: June, 2006 (DJL)
|
Drop me a line