LAN Technologies

Ethernet and IEEE 802.3

This section discusses the development and evolution of Ethernet technologies as defined by both individual members of the industry and the IEEE.

The Evolution of Ethernet

Xerox Corporation usually receives credit for inventing Ethernet. However, Xerox actually acquired the original technology (then known as Aloha Net) in the 1970s from the University of Hawaii. Xerox then joined with DEC and Intel to develop the earliest Ethernet standard, called Version 1, which it released in 1980. The three companies released a follow-up standard, Ethernet Version 2, in 1982.

In the mid-1980s, the IEEE 802 committee adopted Ethernet as the 802.3 standard. All current and future development on Ethernet technologies ostensibly builds on this base standard.

Since its inception, Ethernet has become the most popular LAN standard used throughout the world.

Ethernet Versus IEEE 802.3

It is important to note that Ethernet is not the same as the IEEE 802.3 implementations, and the terms should not be used interchangeably (although they sometimes are). Whereas Xerox, DEC, and Intel developed Version 1 and Version 2 with somewhat similar parameters, the IEEE committee added several features that gave its standard expanded capabilities not shared with its predecessors.

Table 3.1 provides an overview of the similarities and differences between the three implementations.

Table 3.1 Ethernet Versions 1, 2, and IEEE 802.3

Figures 3.1 and 3.2 show examples of unbalanced and balanced signaling.

Figure 3.1 Unbalanced signaling varies voltage levels between 0 (referenced by ground) and +5 volts to represent data.

Figure 3.2 Balanced signaling uses positive and negative voltage levels to represent data.

The unbalanced signaling method used in Version 1 of the Ethernet specification simply uses the presence or lack of voltage to represent data. This type of signaling makes transmissions highly susceptible to outside interference. Ethernet Version 2 improved the signaling method by implementing balanced signaling, representing data through positive and negative voltage changes using 0 or ground as a common reference point. This approach diminishes the effects of interference on transmissions, improving signal quality.

General Ethernet Operation

The IEEE 802.3 specification defines the general operation, components, and distance limitations of Ethernet. They are as follows:

• Defines all Data Link and Physical layer components, functions, channel access method, and operations.

• Provides vendors with rules to follow when implementing or developing Ethernet 802.3 LAN technologies.

• Is based on the IEEE standard known as 10Base5, which all other 802.3 standards follow with minor variations.

The IEEE 802.3 standard defines a 10Mbps broadcast-based linear network architecture using a contention channel access method known as CSMA/CD.

Channel Access Method

Various channel access methods are in use today depending on the network architecture. Ethernet uses a contention-based channel access method. Channel access methods describe the rules used by devices that dictate how the communication medium is accessed, how frames are transmitted, and how the channel is released for use by other devices.

Devices using the CSMA/CD channel access method do the following:

• Contend for the right to transmit

• Can successfully transmit only one at a time

• Must wait for channel availability to transmit a frame when other devices are using the channel (half-duplex operation)

When devices transmit simultaneously on the same channel, signal collisions occur and frames become corrupted.

This contention-based access is called Carrier Sense Multiple Access with Collision Detection (CSMA/CD). Because Ethernet uses silence as the indication to transmit, devices perform a carrier sense to detect that silence. If no frequency exists on the wire at that moment, they can access the channel and begin transmission at once.

After transmission, devices must release the channel and wait at least 9.6µ (microseconds) before attempting to access the channel again, thereby giving other transceivers a chance to transmit their frames.

Collisions

Collisions are just that—collisions. In a baseband network, more than one signal should not be occupying the channel at any one time. The result of more than one signal traversing the wire simultaneously is a collision, which impedes successful transmission (see Figure 3.3).

Figure 3.3 Collision handling (CSMA/CD).

During transmissions, a transceiver (transmitter/receiver) is responsible for encoding the signal on the medium and listening for collisions. If one occurs, the transceiver's internal collision detection circuitry notifies the network adapter card by sending a signal, causing the adapter to abort its transmission.

When devices recognize a collision

• They should stop transmitting, immediately send a 32-bit jamming signal warning other devices of the collision, and then back off for a period of time. Each device will calculate their own back-off algorithm and use the results to determine how long to wait before trying to access the channel again.

• They can attempt to transmit again when the channel is silent and their back-off timer has expired.

It is the responsibility of the transmitting device to detect and retransmit frames when collisions occur.

Collisions are a fact of life with Ethernet, but excessive collisions or late collisions are cause for concern. Overloading a segment with too many devices causes excessive collisions. When too many devices are attached to a segment and each one is contending for the channel, the chance of collisions increases due to sheer volume.

Late collisions are defined in the 802.3 specification as those occurring any time after the 64th byte in a frame. Late collisions can be caused by exceeding maximum distance limitations of the media (known as propagation delay) or by hardware failure, such as a faulty transceiver (also referred to as a jabbering device). You should never consider late collisions a part of normal Ethernet operation and should always investigate them.

Ethernet Frames

Four different frame types exist within the realm of Ethernet standards, each designed with a different purpose by a different entity. The four frame types are as follows:

• Ethernet_II (DIX)

• Ethernet_802.3 (Novell proprietary)

• IEEE 802.3

• IEEE 802.3 SNAP (SubNetwork Access Protocol)

These four frame types can be used on an Ethernet network. The original Ethernet frame known as Ethernet_II was developed by DEC, Intel, and Xerox, which is why this frame is sometimes also referred to as DIX. Novell developed its own proprietary frame (Ethernet_802.3) to be used exclusively for IPX/SPX traffic, and the IEEE developed and named the last two frames.

Even though the developers of these frames assigned specific names to them, de facto naming conventions used within the industry are different. The de facto names are generally recognized and used across all operating system platforms, such as Microsoft's NT and Novell's NetWare. In addition, Cisco has adopted its own naming convention when referring to or configuring these frame types.

Table 3.2 shows the four Ethernet frame types and compares the naming conventions used by the IEEE, industry, and Cisco.

Table 3.2 Ethernet Name Mapping

Table 3.3 contains information specific to each frame type.

Table 3.3 Ethernet Frame Types

All four frame types can coexist in a single network but are not compatible. Only devices using similar frames—or, as Cisco calls them, encapsulations—can communicate with one another.

When stations configured with dissimilar encapsulation types want to exchange information, they must communicate through a router that supports both types. The router performs the conversion between the hosts. However, conversion adds unnecessary overhead and delays to the network, so it's best to select and use only one frame type for your network.

Table 3.4 Similarities of Ethernet Frame Types

Figures 3.4–3.7 illustrate each type of frame. From these, you can compare both functional representations and actual appearances of the frames. Note that all these frames have the same basic characteristics. They all begin with a 6-byte destination MAC address followed by a 6-byte source address. In addition, they all end with a 4-byte CRC field.

Identifying the differences between frames is easier when you use a process of elimination. We know the only frame that uses a protocol type field following the source address is the Ethernet_II or ARPA frame, which makes it easy to isolate it from the others.

Figure 3.4 The Ethernet_II or Cisco (ARPA) frame is the only frame that includes a 2-byte Ether-type value following the source address used to identify the protocol being carried within the frame.

Moving on to the other three frames, you can see that the first three fields are the same (destination address, source address, and length). No distinguishing factors can be seen yet, so you must look further into the frame. Ethernet_802.3 is the Novell proprietary frame type that supports only IPX traffic. It does not use 802.2 SAP addresses in an LLC Header, so it is sometimes referred to as the RAW frame. Always look for an IPX header immediately following the source MAC address when identifying this frame.

Figure 3.5 Ethernet_802.3 (or Novell-Ether) frames always follow the length field with an IPX header that includes a 2-byte null checksum of FFFF.

Figure 3.6 The IEEE's 802.3 frame, known industrywide as Ethernet_802.2 and referred to by Cisco as SAP, includes the 802.2 header with DSAP (destination SAP) and SSAP (source SAP) addresses for protocol identification.

Figure 3.7 The IEEE's 802.3 SNAP frame, known within the industry as Ethernet_SNAP and recognized by Cisco as SNAP, adds a 5-byte extension header (SNAP) to the 802.2 header. This header includes a 3-byte vendor code followed by a 2-byte Ether-type identifying the protocol being carried.

To clarify the differences between Ethernet frame types, Table 3.5 shows individual facets unique to each type.

Table 3.5 Identifying Ethernet Frame Types

Figure 3.8 provides a quick reference comparison of the four frame types discussed previously.

Figure 3.8 Use the Frame Type Quick Reference to quickly identify the differences between each frame.

Half-Duplex Versus Full-Duplex

The term half-duplex communication refers to a mode of transmission in which only one device can be transmitting at a time. This is the basic foundation for Ethernet communication. Devices determine whether the channel is idle before transmitting across their transmit pair, while listening for collisions on their receive pair.

Full-duplex mode enables devices to transmit and receive simultaneously, without listening. Full-duplex capability requires point-to-point connections between devices, such as a switch-to-switch connection. Both devices must support and be configured for full-duplex mode. Full-duplex devices transmit and receive at the same time across a dedicated link that is collision free, which provides the benefit of increased bandwidth capacity and throughput.

Slow Ethernet Specifications and Limitations

Slow (10Mbps) Ethernet has been the mainstay of LAN networks since it came out in the mid-1980s. In the sense that the early days of networking were somewhat chaotic, with vendors making strictly proprietary products and the rules that governed them, it is interesting to follow the development toward de facto industry standards and recent improvements in them. Despite the emergence of standards-defining bodies to provide clear rules for implementing technologies (such as 802.3), desire to exceed limitations drives the industry to ignore many of those rules.

Slow Ethernet specifications include the following:

• 10Base5

• 10Base2

• 10BaseT

10Base5 (Thicknet)

10Base5 connections have the following parameters (as shown in Figure 3.9):

• Transmission takes place at 10Mbps using thick coaxial cable on a linear bus.

• External transceivers attach directly to the medium through vampire taps.

• Tap placements should be at exactly 2.5m or multiples thereof.

• Devices attach to taps using Attachment Unit Interface (AUI) cables connecting to the network card's 15-pin DIX interface.

• Each segment should not exceed 500m; segment extension takes place through repeating devices that provide signal retiming and amplification.

Figure 3.9 10Base5 Thick Ethernet uses coaxial cable as opposed to UTP (Unshielded Twisted Pair), which is commonly used today.

5-4-3 Rule of 10Base5

This rule states that two communicating devices should

• Not be separated by more than five maximum length segments and should have the following:

  • Fifty-ohm terminators at both ends of each segment, with one end connected to a common ground reference
  • A maximum of 100 node attachments per segment
  • A maximum AUI lobe cable distance of 50m



• Not pass through more than four repeaters

Of the five segments, only three should be populated (the remaining two are used strictly to add distance to the overall network). See Figure 3.10.

Figure 3.10 The 5-4-3 Rule states that communicating devices should not be separated by more that five segments connected through four repeaters, with three segments supporting end host connections, while the remaining links are used as unpopulated LAN extensions.

10Base2 (Thinnet)

10Base2 connections have the following parameters (as shown in Figure 3.11):

• Transceivers are generally internal to the network card.

• Transceivers attach to a physical layer thin coaxial (RG-58) bus through BNC connectors.

• Maximum segment distance for RG-58 is 185m.

• Repeaters provide segment extension.

• Segment termination and ground requirements are the same as 10Base5.

• Maximum number of node attachments per segment is 30, with a minimum separation of .5m between each.

• Is also subject to the 5-4-3 Rule.

10BaseT

10BaseT connections have the following parameters (as shown in Figure 3.12):

• Transmit 10Mbps over twisted pairs in a physical star configuration.

• Use two pairs—Pins 1 and 2 for transmitting; pins 3 and 6 for receiving.

• Devices connect to central hubs (known as concentrators) through unshielded twisted pair (UTP) cables with RJ45 connectors.

• Hub-to-hub connections extend the network; the connections require UTP crossover cables unless the conversion is internal.

Figure 3.11 10Base2 Thin Ethernet has a shorter maximum segment length than 10Base5.

• Maximum of 100m cable distance for node-to-hub or hub-to-hub connections.

• Unused hub ports do not require termination.

• Maximum number of node attachments is 1,024.

• Subject only to the 5-4 portion of the 5-4-3 Rule (maximum of five segments through four hubs; each segment can be populated).

Figure 3.12 10BaseT uses UTP cable instead of coaxial cable.

Fast Ethernet Specifications and Limitations

The base standards of Fast Ethernet are exactly the same as those of Slow Ethernet. The difference lies in the additional 10 clauses of the IEEE 802.3u addendum, released in 1995. These clauses define the standard for three different 100Mbps implementations known generally as 100BaseX. Table 3.6 illustrates a comparison between the Slow and Fast Ethernet configurations.

Slow and Fast Ethernet

Table 3.6 delineates the comparisons between Slow Ethernet and Fast Ethernet standards. You should especially note the expansions stemming from the 1995 addendum.

Table 3.6 Fast Versus Slow Ethernet Comparison

100BaseX Standards

All three 100BaseX standards define 100Mb baseband technologies over twisted pair or fiber. Distance and limitations vary within each standard based on physical layer characteristics.

100BaseTX

Of the three IEEE 100BaseX standards, 100BaseTX most closely resembles Slow Ethernet, thereby providing a natural path for migration. You can implement 100BaseX over the same twisted pair infrastructure being used in 10BaseT implementations as long as it's Category 5 UTP. Category 5 is a cabling standard designed to support 100Mb transfer rates. 100BaseTX uses the same transmit and receives pairs, meaning only network interface cards and hubs must be upgraded to support the 100Mb standard. The existing cable plant does not have to be replaced, saving lots of time and money.

100BaseTX has the following characteristics:

• Defines 100Mbps over a minimum Category 5 UTP, implementing the same two pairs as 10Mbps Ethernet

• Encodes signals using the 4B5B encoding scheme

• Replaces Slow Ethernet transceivers and repeaters that support Manchester encoding with devices capable of supporting the 4B5B scheme, such as a 100BaseX repeater, or both schemes, such as a 10/100 switch

• Limits node-to-hub connections to no longer than 100 meters and hub-to-hub connections to 5 meters, making the entire collision domain approximately 205 meters

• Supports half- and full-duplex operation

100BaseFX

Because 100BaseFX uses fiber-optic media, the primary benefit to implementing FX is the capability of achieving greater signaling distances than twisted-pair media. Within the standard, distance limitations depend on such factors as fiber grade (plastic or glass core) and light source—LED (Light Emitting Diode) versus ILD (Injection Laser Diode).

100BaseFX is characterized by the following:

• Defines 100Mbps transfers over fiber-optic media

• Enables half-duplex operation over multi-mode fiber to a maximum overall distance of 412 meters

• Enables full-duplex multi-mode operations up to a maximum of 2 kilometers (1.2 miles)

• Enables full-duplex single-mode fiber operations up to tens of kilometers

• Implements 4B5B signal encoding

100BaseT4

The 100BaseT4 name stems from its requirement of four twisted pairs to function. It uses a different signal encoding scheme than the other 100BaseX standards, making signal conversion necessary when connecting to these other standards. It is primarily used in networks in which older cabling standards have been implemented, such as Category 3 UTP. Category 3 UTP is the first cabling standard to support both voice and data and is capable of supporting 100Mbps, but is not the optimal choice for 100Mb standards.

100BaseT4 is characterized by the following:

• Uses three pairs for transmission and reception and one pair for collision detection

• Is the only 100Base standard that cannot support full-duplex mode, which enables devices to send and receive simultaneously, because it does not dedicate separate pairs for transmission and reception

• Uses 8B6T encoding (which is different from the other 100Base standards)

• Requires signal conversion between itself and any other 100Base standard because it uses a different encoding scheme

• Supports UTP Categories 3, 4, and 5; support of Category 3 enables current Category 3 implementations to migrate to 100Mbps without having to redesign their cabling infrastructure

Class I and Class II Repeaters

100BaseX defines two types of repeaters used to extend the distance of the network. Table 3.7 compares Class I and Class II repeaters, and Figures 3.13 and 3.14 provide graphical presentations of the characteristics of both classes of repeaters.

Table 3.7 Class I and Class II Repeaters

Figure 3.13 Class I repeaters.

Figure 3.14 Class II repeaters.

100VGanyLAN

100VGanyLAN is a Hewlett-Packard proprietary LAN standard meaning Voice Grade over any LAN, signifying that it can support various existing LAN framing methods such as Ethernet or Token-Ring frames. This LAN technology has the following characteristics:

• Offers 100Mbps transmission using a demand priority channel access method

• Is entirely separate from the IEEE 100Base standards

• Requires proprietary hardware and software to implement it

Gigabit Ethernet

Although Gigabit Ethernet technology is just beginning to emerge, it is likely that it will eventually see wide deployment. Its advantages include

• Cost efficiency and savings

• Simplicity of implementation

• Nearly seamless integration with existing Ethernet deployments due to compatibility with existing 10BaseT cabling infrastructures

Gigabit Ethernet Overview

The Gigabit Ethernet standard enables transmission speeds of up to 1000Mbps using Category 5 UTP cabling.

The task force specification is the IEEE 802.3z, which uses 802.3 Ethernet frame formats, as well as the CSMA/CD access method. Note that the continuing use of the 802.3 standard supports backward compatibility with the 100BaseT and 10BaseT technologies.

Table 3.8 shows the cable types and their respective distance limitations on which Gigabit Ethernet is designed to run.

Table 3.8 Gigabit Ethernet Cable Types and Distance Limits