Before reading this you may find the following tutorials useful:
Protocol Basics, Internet Protocol (IP)

In nineteenth-century physics, the “Ether” was believed to be a weightless, transparent, frictionless, and undetectable substance through which all electromagnetic waves travelled. Early in the twentieth century this concept was scrapped as a bad idea, but the term lives on in a protocol known as “Ethernet.”

The TCP/IP protocol suite prepares packets of data from a computer for their foray into the outside world. It straps on the essential addressing information for the data’s journey, as well as giving error detection and correction capabilities. But IP alone cannot take a data packet out of a computer. This function is usually provided by the data-link layer protocol: Ethernet.

Ethernet is a standard for transmission of information through local area networks (LANs) such as office networks. Every device connected to the LAN has an associated address that is hard-wired to it, in contrast to IP addresses that can be changed through software. This is a six-byte MAC (Medium Access Control) address that will be an inherent feature of any Ethernet card connected to a PC. The Ethernet protocol adds its own framing structure to the IP packets, with a header size of up to 26 bytes. The Ethernet frame structure contains the source and destination MAC addresses, as well as control and additional error-checking functionality if required. The minimum data payload size in the Ethernet frame is 46 bytes, with a maximum of 1500 bytes.

An Ethernet Frame Ethernet now provides access to the physical medium through which the data must travel – the cable connected to the PC. This is achieved through a method known as Carrier Sense Multiple Access/Collision Detection (CSMA/CD). This technique works a lot like crossing the road. You approach the curb and look to see if there is any traffic in the road. In this case, the Ethernet card takes a look out onto the cable and senses any activity in the form of traffic from other PCs on the link. If it senses traffic on the cable, then it waits a set time before looking again. Just as if you see a large truck coming along you will, ideally, wait before trying to cross.

Once the Ethernet card detects no activity on the link, it begins to send out its own Ethernet frames in the form of electrical 1s and 0s. However, there is the chance that another device has also looked at a similar time and sent out its own Ethernet frames. In this case there is a collision and no data gets through. Ethernet is designed to detect this and therefore has the ability to retransmit the information again when it will hope to have more success.

With many PCs all connected via the same cable, the likelihood of collision is so high that it’s surprising any information ever gets transferred at all. It is indeed these collisions that limit the possible length of cable in a system, as greater lengths of cable provide more scope for devices to transmit at similar times without detecting activity on the link. A technology known as switched Ethernet aims to improve matters by dividing links into separate sections, thus reducing the number of devices with access to one particular section and so reducing collisions.

The destination MAC address in the Ethernet frame takes it onto the nearest IP router, which strips off the Ethernet header to handle the raw IP packet. Once the router has decided where to send the packet next, towards its IP destination address, it is again framed by Ethernet and retransmitted using CSMA/CD once more.

Ethernet comes in a variety of flavors. The most often encountered is known as 10Base-T, which has a 10-Mbit/s transfer rate and uses unshielded twisted pair (UTP) cable as its access medium. The length limit on such a system is around 100 meters. 100Base-T is often known as “fast Ethernet” and gives 100-Mbit/s transfer speeds, again on UTP cable with an upper length limit of 100m. The most modern flavor is “gigabit Ethernet,” which provides 1-Gbit/s access rates on optical fiber over distances up to several kilometers.

Key Points

  • Takes IP packets out into the network
  • Uses Medium Access Control (MAC) addresses for devices
  • Ethernet frame carries data along with source and destination MAC addresses and error-checking information
  • Accesses unshielded twisted pair (UTP) cable by sensing activity and only transmitting if nobody else on the cable is doing so
  • Collisions can limit the range of Ethernet networks
  • Fast Ethernet gives speeds of up to 100 Mbit/s over distances up to 100m
  • Gigabit Ethernet can transmit on optical fiber over distances of several kilometers

Further Reading

Formatting for Transmission, Sonet (Synchronous Optical NETwork) and SDH (Synchronous Digital Hierarchy)
COMMENTS Add Comment
Page 1 / 8   >   >>
sraghuna 12/5/2012 | 1:58:21 AM
re: Ethernet Could anyone point me to some literature on PHY working/architecture ? Appreciate your help.
sigint 12/5/2012 | 1:58:19 AM
re: Ethernet Phy vendors like broadcom and Marvell are pretty secretive about their technologies and products. Try googling for LXT1000, a (now) canned intel device, the datasheet had some decent info.

You could also try Cicada Semiconductors, now a part of Vitesse.
Buddy_Lite 12/5/2012 | 12:52:31 AM
re: Ethernet
SendMeLight wrote,

"Is there anyway to configure the Fast ethernet and Gigabit ethernet ports to limit the throughput to less than the maximum allowed? If yes, could you list the vendors who support such a configuration? "

Yes. Go to Http://www.coriolisnet.com
They allow you to run TDM and packet traffic on the same SONET ring.
Packet traffic is policed in 1/2Mb increments.
bscan 12/4/2012 | 11:36:41 PM
re: Ethernet This tutorial states that a destination and source MAC address is sent with the Ethernet Packet. A MAC address is a hardware address that exists on the ethernet card. No way when I send a message out there do I know the MAC address of the destination. I assume we start with an IP (XXX.XXXX.XXX.XXX) address and somehow it is eventually linked to the MAC address of the destination device. Something appears to be missing in the description.
fiber_r_us 12/4/2012 | 11:36:40 PM
re: Ethernet For LAN media (Ethernet, Token Ring, FDDI, etc), a protocol exists called the Address Resolution Protocol (ARP). Its job is to discover the MAC address of the next-hop node in the path towards the final IP destination. It is this MAC address that is placed in the MAC header as the MAC destination address. The ARP process is done independently for each "Layer 2" domain along a path towards the final IP destination.
RJ-45 12/4/2012 | 11:36:39 PM
re: Ethernet Yup, it's missing all right: it's called ARP (Address Resolution Protocol).
walter_100 12/4/2012 | 11:36:37 PM
re: Ethernet bscan,
I always know the IP address of my destination. In case the destn belongs to the same subnet:
1) The MAC address would/could be cached for that particular IP address/netmask.
2) In case not, it asks the default gateway router for the MAC address.
3) In case the Default Gateway doesn't have it, it would do an ARP by broadcasting the query on the entire subnet. The machine with that particular IP address responds with its MAC address.
4) The machine would cache the entry and flush it every 1/2 hr or so....

In case, IP Address of destn belongs to another network, it would just populate the MAC address of the default gateway in its frame.
The router in the detn network will populate it with its correct MAC address.

fiber_r_us 12/4/2012 | 11:36:36 PM
re: Ethernet
>2) In case not, it asks the default gateway
>router for the MAC address.

I am not aware of any case where a system on a given subnet would ask the default gateway for a MAC address for a destination system on the same subnet. The system either:

1) Already has the MAC address of the destination system in cache (due to a previous ARP request or due to listening to ARP requests on that subnet sent by other machines on that subnet (promiscuous ARP))

2) Sends an ARP request to the broadcast MAC address (which should not leave that subnet) with the destination's IP address in the ARP request. The destination system will receive the packet (because it is a broadcast) and will respond directly to the source node.
walter_100 12/4/2012 | 11:36:36 PM
re: Ethernet fiber_r_us,
Sounds reasonable. You should be right.
I stand corrected....
naaman 12/4/2012 | 11:13:14 PM
re: Ethernet Can some one please tell me how to sketch an ethernet frame and what are the functions of the each field in the frame
Page 1 / 8   >   >>
Sign In