Latest update on Thursday 16 October 2008 à 09:43 by Jean-François Pillou.
Introduction to Ethernet
Ethernet (also known as IEEE 802.3 standard) is a data transmission standard for local area networks based on the following principle:
All machines on an Ethernet network
are connected to the same communication line,
made up of cylindrical cables
Different variants of Ethernet technologies are distinguished according to the type and diameter of the cables used:
10Base2: The cable used is a thin coaxial cable, called thin Ethernet,
10Base5: The cable used is a thick coaxial cable, called thick Ethernet,
10Base-T: The cable used is a twisted pair (the T means twisted pair), the speed achieved is around 10 Mbps,
100Base-FX: Makes it possible to achieve a speed of 100Mbps by using multimode fibre optic (F stands for Fibre).
100Base-TX: Like 10Base-T, but with a speed 10 times greater (100Mbps),
1000Base-T: Uses a double twisted pair of category 5 cables and allows a speed of a Gigabit per second.
1000Base-SX: Based on multimode fibre optic using a short wavelength signal (S stands for short) of 850 nanometres (770 to 860 nm).
1000Base-LX: Based on multimode fibre optic using a long wavelength signal (L stands for long) of 1350 nm (1270 to 1355 nm).
Abbreviation
Name
Cable
Connector
Speed
Ports
10Base2
Thin Ethernet
Coaxial cable (50 Ohms) with a thin diameter
BNC
10 Mb/s
185m
10Base5
Thick Ethernet
Coaxial cable with a thick diameter (0.4 inch)
BNC
10Mb/s
500m
10Base-T
Standard Ethernet
Twisted pair (category 3)
RJ-45
10 Mb/s
100m
100Base-TX
Fast Ethernet
Double twisted pair (category 5)
RJ-45
100 Mb/s
100m
100Base-FX
Fast Ethernet
Multimode fibre optic (type 62.5/125)
100 Mb/s
2 km
1000Base-T
Gigabit Ethernet
Double twisted pair (category 5)
RJ-45
1000 Mb/s
100m
1000Base-LX
Gigabit Ethernet
Monomode or multimode fibre optic
1000 Mb/s
550m
1000Base-SX
Gigabit Ethernet
Multimode fibre optic
1000 Mbit/s
550m
10GBase-SR
10Gigabit Ethernet
Multimode fibre optic
10 Gbit/s
500m
10GBase-LX4
10Gigabit Ethernet
Multimode fibre optic
10 Gbit/s
500m
Ethernet is a widely used network technology because the cost of such a network is not very high
Transmission principle
All the computers on an Ethernet network are connected to the same transmission line, and communication is carried out using a protocol called CSMA/CD (Carrier Sense Multiple Access with Collision Detect which means that it is a multiple access protocol with carrier monitoring (Carrier Sense) and collision detection).
With this protocol any machine is authorized to transmit over the line at any time and with no concept of priority between machines. This communication is conducted simply:
Each machine verifies that there is no communication on the line before transmitting
If two machines transmit simultaneously, then there is a collision (i.e. several data frames are located on the line at the same time)
The two machines interrupt their communication and wait for a random time period, then once the first has exceeded this time delay it can then retransmit
This principle is based on several limitations:
The data packets must have a maximum size
there must be a waiting time between two transmissions
The waiting time varies depending on the frequency of collisions:
After the first collision a machine waits for a unit of time
After the second collision the machine waits for two units of time
After the third collision the machine waits for four units of time
... with of course a small amount of additional random time
Switched Ethernet
Until now, the Ethernet topology described has been that of the shared Ethernet (any message transmitted is heard by all connected machines, available bandwidth is shared by all machines).
Over several years a significant development has taken place: that of the Switched Ethernet. The physical topology remains a star, but is organized around a switch. The switch uses a filtering and switching mechanism very similar to that used by gateways where these techniques have been used for a long time.
It inspects the source and destination addresses of messages, draws up a table which then allows it to know which machine is connected on which port of the switch (in general this process is done by self-learning, i.e. automatically, but the switch manager can carry out additional adjustments).
Knowing the recipient port, the switch will only transmit the message on the appropriate port, the other ports remaining free for other transmissions which can be carried out simultaneously. The result is that each exchange can be carried out at nominal speed (greater division of the bandwidth), without collisions, with a noticeable increase in the network bandwidth as a result (also at nominal speed).
As for knowing whether all the ports on a switch can communicate at the same time without loosing messages, that depends on the quality of the switch (non blocking switch).
Since switching makes it possible to avoid collisions and the 10/100/1000 base T(X) technologies have separate circuits for transmission and reception (a twisted pair per transmission direction), the majority of modern switches make it possible to deactivate collision detection and move into full-duplex mode on the ports. In this way, the machines can transmit and receive at the same time (which again contributes to network performance).
Full-duplex mode is particularly interesting for servers which must serve several clients.
Modern Ethernet switches also detect the transmission speed used by each machine (autosensing) and if the machine supports several speeds (10 or 100 or 1000 megabits/sec) begin by negotiating with it to choose a speed as well as the semi-duplex or full-duplex mode of the transmission. This makes it possible to have a store of machines with different performances (for example a set of computers with various hardware configurations).
As the traffic transmitted and received is no longer transmitted over all ports, it becomes much more difficult to sniff what is happening. This contributes to the general security of the network, which is a highly sensitive subject today.
To finish, the use of switches makes it possible to build geographically larger networks. On shared Ethernet, a message must be able to wait for any other machine on the network for a specified time period (slot time) without which the collision detection mechanism (CSMA/CD) does not work correctly.
This no longer applies with Ethernet switches. The distance is no longer limited except by the technical limits of the medium used (fibre optic or twisted pair, strength of signal transmitted and sensitivity of the receiver,...).
