This post is an overview of the most popular LAN technology: Ethernet.
A bit of Ethernet history
At first, the DIX group developed Ethernet II, then IEEE 802.3 group adopted Ethernet and added to it some improvements. All Ethernet technologies after Ethernet II are defined by the IEEE 802.3 body, from the early 10Base-5 and 10Base-2 till today. Standards such as Fast Ethernet and Gigabit Ethernet came to existence. And nowadays we are talking about 10Gigabit Ethernet and even 40Gigabit Ethernet.
IEEE 802.3 is a working group that standardized wired Ethernet, on both layer 1 layer 2 of OSI model, just like other IEEE working groups. Each IEEE working group specializes in one technology. For example, IEEE 802.11 works on WiFi.
All LAN and MAN technologies (such as Broadband, Wimax,…) can be referred to as IEEE 802 LAN/MAN standards.
10Base-T uses category 3 UTP cables that were used for voice telephony. This cabling was abundant in corporate buildings. then with 100Base-TX we saw UTP cables that had two twisted pairs of copper. And with the increase in bandwidth, the sensibility to noise increased on copper cables. So technologies such as Foiled Twisted Pair FTP and Shielded Twisted Pair STP appeared on the market.
Network techies often interchangeably use the terms “LAN cable”, “Ethernet cable” and “Network cable” to mean copper cables with RJ45 connectors. That’s because most LAN cabling schemes use UTP or STP cables -except for respected data centers-.
Ethernet LLC and MAC
Although the OSI model defines the data link layer, IEEE divides the data link layer of the OSI model into two sub-layers:
- the MAC sub-layer
- the LLC sub-layer
IEEE 802.2 LLC provides both connection-oriented services and connection-less services to the Network layer. LLC was designed to allow multiplexing of network layer protocols such as IP and IPX. This means IP and IPX traffic can both be carried over the same Ethernet media.
Also, LLC acts as the transport protocol for the popular Spanning Tree Protocol.
The LLC sub-layer is generally the same for all IEEE 802 LAN and MAN networks. However the MAC sub-layer is different.
Ethernet Frame format
The first format of an Ethernet frame was called Ethernet II, or Ethernet v2. And since the appearance of 10Base-5 specifications, Ethernet II lost its popularity in favor of Ethernet 802.3.
The differences between Ethernet II frame format and Ethernet 802.3 frame format are :
- the Type/Length field. In Ethernet II, we have the Type field (aka EtherType). Its values determines what type of protocol is in the Data field afterwards. In Ethernet 802.3, that same field is called Length.
- The addition of the 802.2 sub-header (LLC) to the IEEE 802.3 header.
When LLC sub-layer is used, the frame is called IEEE 802.3 frame with LLC sub-layer or simply 802.2 LLC frame. The LLC sub-layer -or LLC sub-header- brings with it three sub-fields:
- Destination Service Access Point DSAP
- Source Service Access Point SSAP
For backward compatibility with network devices that solely support Ethernet II frames, a special portion called SNAP sub-frame is added to the IEEE 802.3 Ethernet frame. But there is a condition: that LLC sub-layer is present in the frame.
The SNAP sub-frame is composed of the following two parts:
- OUI: Organizationally Unique Identifier
- Type: provides the same values as in Ethernet II frames.
So we have the following types of Ethernet frame formats:
- the old Ethernet II
- IEEE 802.3
- IEEE 802.3 with LLC sub-layer, aka IEEE 802.2 LLC frame
- IEEE 802.3 with LLC/SNAP
By the way, we must distinguish between what is called the IEEE packet and the Ethernet frame. The IEEE packet includes the Ethernet 802.3 frame plus the Preamble and a trailing code.
Ethernet and inter-frame gap
Between one Ethernet frame and another, there is a gap called inter-frame gap, whose worth 12 bytes (equivalent to 9.6 micro-seconds). This gap helps to distinguish frames.
Ethernet and frame rate
Frame rate is the number of frames produced per seconds for a specific data rate. It is equal to the data rate divided by the size of the frame.
Frame rate = data rate / frame size
If we take 10Base-T Ethernet, for an Ethernet frame of 1518 bytes, we get a frame rate of 10*10^6/(1518*8) = 823 frames per second.
Link layer throughput or efficiency
The throughput of a link, or the efficiency of a link, is measured by multiplying the frame rate by the size of the frame payload.
Link layer throughput = frame rate * frame payload
For example, given the frame rate of the example above, and a paylod of 1500 bytes, we get a layer 2 efficiency of:
823 * 1500*8 = 9876000 bps, or 9.8 Mbps. This gives a ratio of 9876000/10000000 = 98.7%
In a previous post, we said that MTU is the maximum size of the payload inside a layer 2 technology. Let’s focus on Ethernet MTU and on IP encapsulated inside an Ethernet frame.
The MTU includes the IP packet (payload and headers) but it does not include link layer encapsulation headers. So for example if we have IP over Ethernet, in addition to the 1500 bytes, an Ethernet frame adds 18 bytes:
- destination MAC: 6 Bytes
- source MAC: 6 Bytes
- Type/Length: 2 Bytes
- FCS: 4 bytes
- the preamble and the SFD are not counted as part of an Ethernet frame.
–> The total frame size becomes 1500 + 6 + 6 + 2 + 4 = 1518 bytes.
Ethernet and CSMA/CD formula
Remember the formula that determines the minimum packet size in CSMA/CD networks? CSMA/CD is still implemented in modern Ethernet networks. However, we no longer want to restrict the size of the Ethernet frame. Payloads from the upper layer protocols can have considerable sizes. So the final decision made by the early super network engineers and architects was to reduce the distance L in favor of the data rate and the packet size. This is where the maximum allowed distance of 100m in switched Ethernet networks comes from.
Having a maximum length of the switched Ethernet segment of 100 meters means there won’t be any collisions on the medium from end to end. Above the 100 meters, there is no guarantee for the quality of the signal transmission.
But where does the 100m begin and where does it end?
A standard horizontal cabling scheme involves a computer, patch cords, a jack, a patch panel and a switch. The rule is that the Ethernet cable from the computer to the first Ethernet switch must not exceed 100m.
What if a computer from one building desires to connect with anoter host in another building? Does the 100m switched Ethernet distance condition holds true?
In switched Ethernet networks, the maximum distance of 100m is true for each network segment between a host and the switch port, and between switches together (uplink). So:
- in building 3, an Ethernet cable from a host to the switch port must not exceed 100m, including patch cords at the wall jacket and patch cords at the patch panel
- between buildings 1 and 3, the Ethernet cable length must not exceed 100m. Otherwise, it must be changed with fiber optic cable.
- in building 1, the same rule apply as in 1.
- Introduction to Computer Networking, Stanford University
- TCP/IP Illustrated: The Protocols
- Upgrading And Repairing Networks
- CCIE Routing And Switching Exam Guide 3rd Edition
Figure 1: Cisco.com
Figure 2: TCP/IP Illustrated, Vol 1: The Protocols
Figure 3: Upgrading And Repairing Networks
Figure 4: CCIE Routing And Switching Exam Guide, 3rd Edition
Figure 5: FlukeNetworks.com