BUY THIS BOOK
Add to Cart

PDF $35.99

Safari Books Online

What is this?

Looking to Reprint or License this content?


Ethernet: The Definitive Guide
Ethernet: The Definitive Guide By Charles E. Spurgeon
February 2000
Pages: 520

Cover | Table of Contents | Colophon


Table of Contents

Chapter 1: The Evolution of Ethernet
Ethernet is by far the most widely used local area networking (LAN) technology in the world today. Market surveys indicate that hundreds of millions of Ethernet network interface cards (NICs), repeater ports, and switching hub ports have been sold to date, and the market continues to grow. In total, Ethernet outsells all other LAN technologies by a very large margin.
Ethernet reached its 25th birthday in 1998, and has seen many changes as computer technology evolved over the years. Ethernet has been constantly reinvented, evolving new capabilities and in the process growing to become the most popular network technology in the world.
This chapter describes the invention of Ethernet, and the development and organization of the Ethernet standard. Along the way we provide a brief tour of the entire set of Ethernet media systems.
On May 22, 1973, Bob Metcalfe (then at the Xerox Palo Alto Research Center, PARC, in California) wrote a memo describing the Ethernet network system he had invented for interconnecting advanced computer workstations, making it possible to send data to one another and to high-speed laser printers. Probably the best-known invention at Xerox PARC was the first personal computer workstation with graphical user interfaces and mouse pointing device, called the Xerox Alto. The PARC inventions also included the first laser printers for personal computers, and, with the creation of Ethernet, the first high-speed LAN technology to link everything together.
This was a remarkable computing environment for the time, since the early 1970s were an era in which computing was dominated by large and very expensive mainframe computers. Few places could afford to buy and support mainframes, and few people knew how to use them. The inventions at Xerox PARC helped bring about a revolutionary change in the world of computing.
A major part of this revolutionary change in the use of computers has been the use of Ethernet LANs to enable communication among computers. Combined with an explosive increase in the use of information sharing applications such as the World Wide Web, this new model of computing has brought an entire new world of communications technology into existence. These days, sharing information is most often done over an Ethernet; from the smallest office to the largest corporation, from the single schoolroom to the largest university campus, Ethernet is clearly the networking technology of choice.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
History of Ethernet
On May 22, 1973, Bob Metcalfe (then at the Xerox Palo Alto Research Center, PARC, in California) wrote a memo describing the Ethernet network system he had invented for interconnecting advanced computer workstations, making it possible to send data to one another and to high-speed laser printers. Probably the best-known invention at Xerox PARC was the first personal computer workstation with graphical user interfaces and mouse pointing device, called the Xerox Alto. The PARC inventions also included the first laser printers for personal computers, and, with the creation of Ethernet, the first high-speed LAN technology to link everything together.
This was a remarkable computing environment for the time, since the early 1970s were an era in which computing was dominated by large and very expensive mainframe computers. Few places could afford to buy and support mainframes, and few people knew how to use them. The inventions at Xerox PARC helped bring about a revolutionary change in the world of computing.
A major part of this revolutionary change in the use of computers has been the use of Ethernet LANs to enable communication among computers. Combined with an explosive increase in the use of information sharing applications such as the World Wide Web, this new model of computing has brought an entire new world of communications technology into existence. These days, sharing information is most often done over an Ethernet; from the smallest office to the largest corporation, from the single schoolroom to the largest university campus, Ethernet is clearly the networking technology of choice.
Bob Metcalfe's 1973 Ethernet memo describes a networking system based on an earlier experiment in networking called the Aloha network. The Aloha network began at the University of Hawaii in the late 1960s when Norman Abramson and his colleagues developed a radio network for communication among the Hawaiian Islands. This system was an early experiment in the development of mechanisms for sharing a common communications channel—in this case, a common radio channel.
The Aloha protocol was very simple: an Aloha station could send whenever it liked, and then waited for an acknowledgment. If an acknowledgment wasn't received within a short amount of time, the station assumed that another station had also transmitted simultaneously, causing a
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
The Latest Ethernet Standard
After the publication of the original IEEE 802.3 standard for thick Ethernet, the next development in Ethernet media was the thin coaxial Ethernet variety, inspired by technology first marketed by the 3Com Corporation. When the IEEE 802.3 committee standardized the thin Ethernet technology, they gave it the shorthand identifier of 10BASE2, which is explained later in this chapter.
Following the development of thin coaxial Ethernet came several new media varieties, including the twisted-pair and fiber optic varieties for the 10 Mbps system. Next, the 100 Mbps Fast Ethernet system was developed, which also included several varieties of twisted-pair and fiber optic media systems. Most recently, the Gigabit Ethernet system was developed using both fiber optic and twisted-pair cabling. These systems were all developed as supplements to the IEEE Ethernet standard.
When the Ethernet standard needs to be changed to add a new media system or capability, the IEEE issues a supplement which contains one or more sections, or "clauses" in IEEE-speak. The supplement may consist of one or more entirely new clauses, and may also contain changes to existing clauses in the standard. New supplements to the standard are evaluated by the engineering experts at various IEEE meetings and the supplements must pass a balloting procedure before being voted into the full standard.
New supplements are given a letter designation when they are created. Once the supplement has completed the standardization process, it becomes part of the base standard and is no longer published as a separate supplementary document. On the other hand, you will sometimes see trade literature that refers to Ethernet equipment with the letter of the supplement in which the variety was first developed (e.g., IEEE 802.3u may be used as a reference for Fast Ethernet). Table 1.1 lists several supplements and what they refer to. The dates indicate when formal acceptance of the supplement into the standard occurred. Access to the complete set of supplements is provided in Appendix A.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Organization of IEEE Standards
The IEEE standards are organized according to the Open Systems Interconnection (OSI) Reference Model. This model was developed in 1978 by the International Organization for Standardization, whose initials (derived from its French name) are ISO. Headquartered in Geneva, Switzerland, the ISO is responsible for setting open, vendor-neutral standards and specifications for items of technical importance. For example, if you're a photographer you've no doubt noticed the ISO standard speeds for camera film.
The ISO developed the OSI reference model to provide a common organizational scheme for network standardization efforts (with perhaps an additional goal of keeping us all confused with reversible acronyms). What follows is a quick, and necessarily incomplete, introduction to the subject of network models and international standardization efforts.
The OSI reference model is a method of describing how the interlocking sets of networking hardware and software can be organized to work together in the networking world. In effect, the OSI model provides a way to arbitrarily divide the task of networking into separate chunks, which are then subjected to the formal process of standardization.
To do this, the OSI reference model describes seven layers of networking functions, as illustrated in Figure 1.2. The lower layers cover the standards that describe how a LAN system moves bits around. The higher layers deal with more abstract notions, such as the reliability of data transmission and how data is represented to the user. The layers of interest for Ethernet are the lower two layers of the OSI model.
Figure 1.2: The OSI seven layer model
In brief, the OSI reference model includes the following seven layers, starting at the bottom and working our way to the topmost layer:
Physical layer
Standardizes the electrical, mechanical, and functional control of data circuits that connect to physical media.
Data link layer
Establishes communication from station to station across a single link. This is the layer that transmits and receives frames, recognizes link addresses, etc. The part of the standard that describes the Ethernet frame format and MAC protocol belongs to this layer.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Levels of Compliance
In developing a technical standard, the IEEE is careful to include only those items whose behavior must be carefully specified to make the system work. Therefore, all Ethernet interfaces that operate in the original half-duplex mode (described in Chapter 3) must comply fully with the MAC protocol specifications in the standard to perform the functions identically. Otherwise, the network would not function correctly.
At the same time, the IEEE makes an effort not to constrain the market by standardizing such things as the appearance of an Ethernet interface, or how many connectors it should have on it. The intent is to provide just enough engineering specifications to make the system work reliably, without inhibiting competition and the inventiveness of the marketplace. In general, the IEEE has been quite successful. Most equipment designed for use in an Ethernet system fully complies with the standard.
Vendor innovation can sometimes lead to the development of devices that are not described in the IEEE standard, and that are not included in the half-duplex mode timing specs or the media specs in the standard. Some of these devices may work well for a small network, but might cause problems with signal timing in a larger network operating in half-duplex mode. Further, a network system using equipment not described in the standard or included in the official guidelines cannot be evaluated using the IEEE half-duplex mode configuration guidelines.
How much you should be concerned about all this is largely up to you and your particular circumstances. Another way of saying this is: "Optimality differs according to context." It's up to you to decide how important these issues are, given your particular circumstances (or context). For one thing, not all innovations are a bad idea.
After all, the thin coaxial and twisted-pair Ethernet media systems started life as vendor innovations that later became carefully specified media systems in the IEEE standard. However, if your goal is maximum predictability and stability for your network given a variety of vendor equipment and traffic loads, then one way to help achieve that goal is by using only equipment that is described in the standard.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
IEEE Identifiers
The IEEE has assigned shorthand identifiers to the various Ethernet media systems as they have been developed. The three-part identifiers include the speed, the type of signaling used, and information about the physical medium.
In the early media systems, the physical medium part of the identifier was based on the cable distance in meters, rounded to the nearest 100 meters. In the more recent media systems, the IEEE engineers dropped the distance convention and the third part of the identifier simply identifies the media type used ( twisted-pair or fiber optic). In roughly chronological order, the identifiers include the following set:
10BASE5
This identifies the original Ethernet system, based on thick coaxial cable. The identifier means 10 megabits per second transmission speed, base band transmission, and the 5 refers to the 500 meter maximum segment length. The word baseband simply means that the transmission medium, thick coaxial cable in this instance, is dedicated to carrying one service: Ethernet signals. The 500 meter limit refers to the maximum length a given cable segment may be. Longer networks are built by connecting multiple segments with repeaters or switching hubs.
10BASE2
Also known as the thin Ethernet system, this media variety operates at 10 Mbps, in baseband mode, with cable segment lengths that can be a maximum of 185 meters in length. If the segments can be at most 185 meters long, then why does the identifier say "2," thus implying a maximum of 200 meters? The answer is that the identifier is merely a bit of shorthand and not intended to be an official specification. The IEEE committee found it convenient to round things up to 2, to keep the identifier short and easier to pronounce. This less expensive version of coax Ethernet was nicknamed "Cheapernet."
FOIRL
This stands for Fiber Optic Inter-Repeater Link . The original DIX Ethernet standard mentioned a point-to-point link segment that could be used between repeaters, but did not provide any media specifications. Later, the IEEE committee developed the FOIRL standard, and published it in 1989. FOIRL segments were originally designed to link remote Ethernet segments together. Fiber optic media's immunity to lightning strikes and electrical interference, as well as its ability to carry signals for long distances, makes it an ideal system for transmitting signals between buildings.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Reinventing Ethernet
No matter how well designed a LAN system is, it won't help you much if you can only use it with a single vendor's equipment. A LAN has to be able to work with the widest variety of equipment possible to provide you with the greatest flexibility. For maximum utility, your LAN must be vendor-neutral: that is, capable of interworking with all types of computers without being vendor-specific. This was not the way things worked in the 1970s when computers were expensive and networking technology was exotic and proprietary.
Bob Metcalfe understood that a revolution in computer communications required a networking technology that everyone could use. In 1979 he set out to make Ethernet an open standard, and convinced Xerox to join a multi-vendor consortium for the purposes of standardizing an Ethernet system that any company could use. The era of open computer communications based on Ethernet technology formally began in 1980 when the Digital Equipment Corporation (DEC), Intel, and Xerox consortium announced the DIX standard for 10 Mbps Ethernet.
This DIX standard made the technology available to anyone who wanted to use it, producing an open system. As part of this effort, Xerox agreed to license its patented technology for a low fee to anyone who wanted it. In 1982 Xerox also gave up its trademark on the Ethernet name. As a result, the Ethernet standard became the world's first open, multi-vendor LAN standard. The idea of sharing proprietary computer technology in order to arrive at a common standard to benefit everyone was a radical notion for the computer industry in the late 1970s. It's a tribute to Bob Metcalfe's vision that he realized the importance of making Ethernet an open standard. As Metcalfe put it:
The invention of Ethernet as an open, non-proprietary, industry-standard local network was perhaps even more significant than the invention of Ethernet technology itself.
In 1979 Metcalfe started a company to help commercialize Ethernet. He believed that computers from multiple vendors ought to be able to communicate compatibly over a common networking technology, making them more useful and, in turn, opening up a vast new set of capabilities for the users.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Multi-Gigabit Ethernet
In March 1999, the IEEE 802.3 standards group held a "Call for Interest" meeting on the topic of Ethernet speeds beyond the current 1 Gbps standard. A number of presentations were made on the general topic, after which the group voted to create a High Speed Study Group. At the time of this writing, the High Speed Study Group is meeting on a regular basis to review presentations on a variety of technical issues. It is expected that this work will lead to the development of a new higher speed Ethernet standard operating at 10 Gbps within the next few years.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Chapter 2: The Ethernet System
An Ethernet Local Area Network (LAN) is made up of hardware and software working together to deliver digital data between computers. To accomplish this task, four basic elements are combined to make an Ethernet system. This chapter provides a tutorial describing these elements, since a familiarity with these basic elements provides a good background for working with Ethernet. We will also take a look at some network media and simple topologies. Finally, we will see how the Ethernet system is used by high-level network protocols to send data between computers.
This chapter describes the original half-duplex mode of operation. Half-duplex simply means that only one computer can send data over the Ethernet channel at any given time. In half-duplex mode, multiple computers share a single Ethernet channel by using the Carrier Sense Multiple Access with Collision Detection (CSMA/CD) media access control (MAC) protocol. Until the introduction of switching hubs, the half-duplex system was the typical mode of operation for the vast majority of Ethernet LANs—tens of millions of Ethernet connections have been installed based on this system.
However, these days many computers are connected directly to their own port on an Ethernet switching hub and do not share the Ethernet channel with other systems. This type of connection is described in Chapter 18. Many computers and switching hub connections now use full-duplex mode, in which the CSMA/CD protocol is shut off and the two devices on the link can send data whenever they like. The full-duplex mode of operation is described in Chapter 4.
The Ethernet system includes four building blocks that, when combined, make a working Ethernet:
  • The frame , which is a standardized set of bits used to carry data over the system.
  • The media access control protocol , which consists of a set of rules embedded in each Ethernet interface that allow multiple computers to access the shared Ethernet channel in a fair manner.
  • The signaling components
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Four Basic Elements of Ethernet
The Ethernet system includes four building blocks that, when combined, make a working Ethernet:
  • The frame , which is a standardized set of bits used to carry data over the system.
  • The media access control protocol , which consists of a set of rules embedded in each Ethernet interface that allow multiple computers to access the shared Ethernet channel in a fair manner.
  • The signaling components , which consist of standardized electronic devices that send and receive signals over an Ethernet channel.
  • The physical medium , which consists of the cables and other hardware used to carry the digital Ethernet signals between computers attached to the network.
The heart of the Ethernet system is the frame. The network hardware—which is comprised of the Ethernet interfaces, media cables, etc.—exists simply to move Ethernet frames between computers, or stations. The bits in the Ethernet frame are formed up in specified fields. Figure 2.1 shows the basic frame fields. These fields are described in more detail in Chapter 3.
Figure 2.1: An Ethernet frame
Figure 2.1 shows the basic Ethernet frame, which begins with a set of 64 bits called the preamble . The preamble gives all of the hardware and electronics in a 10 Mbps Ethernet system some signal start-up time to recognize that a frame is being transmitted, alerting it to start receiving the data. This is what a 10 Mbps network uses to clear its throat, so to speak. Newer Ethernet systems running at 100 and 1000 Mbps use constant signaling, which avoids the need for a preamble. However, the preamble is still transmitted in these systems to avoid making any changes in the structure of the frame.
Following the preamble are the destination and source addresses. Assignment of addresses is controlled by the IEEE Standards Association (IEEE-SA), which administers a portion of the address field. When assigning blocks of addresses for use by network vendors, the IEEE-SA provides a 24-bit Organizationally Unique Identifier (OUI). The OUI is a unique 24-bit identifier assigned to each organization that builds network interfaces. This allows a vendor of Ethernet equipment to provide a unique address for each interface they build. Providing unique addresses during manufacturing avoids the problem of two or more Ethernet interfaces in a network having the same address. This also eliminates any need to locally administer and manage Ethernet addresses.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Ethernet Hardware
The next two building blocks of an Ethernet system include the hardware components used in the system. There are two basic groups of hardware components: the signaling components, used to send and receive signals over the physical medium; and the media components, used to build the physical medium that carries the Ethernet signals. Not surprisingly, these hardware components differ depending on the speed of the Ethernet system and the type of cabling used. To show the hardware building blocks, we'll look at an example based on the widely used 10 Mbps twisted-pair Ethernet media system, called 10BASE-T.
The signaling components for a twisted-pair system include the Ethernet interface located in the computer, as well as a transceiver and its cable. An Ethernet may consist of a pair of stations linked with a single twisted-pair segment, or multiple stations connected to twisted-pair segments that are linked together with an Ethernet repeater. A repeater is a device used to repeat network signals onto multiple segments. Connecting cable segments with a repeater makes it possible for the segments to all work together as a single shared Ethernet channel.
Figure 2.2 shows two computers (stations ) connected to a 10BASE-T media system. Both computers have an Ethernet interface card installed, which makes the Ethernet system operate. The interface contains the electronics needed to form up and send Ethernet frames, as well as to receive frames and extract the data from them. The Ethernet interface comes in two basic types. The first is a board that plugs into a computer's bus slot, and the other relies on chips that allow Ethernet interfaces to be built into the computer's main logic board. In the second form, all you'll see of the interface is an Ethernet connector mounted on the back of the computer.
Figure 2.2: A sample 10BASE-T Ethernet connection
The Ethernet interface connects to the media system using a transceiver, which can be built into the interface or provided as an external device. Of the two stations shown in Figure 2.2, one is provided with a built-in transceiver and one uses an external transceiver. The word "transceiver" is a combination of
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Network Protocols and Ethernet
Now that we've seen how frames are sent over Ethernet systems, let's look at the data being carried by the frame. Data that is being sent between computers is carried in the data field of the Ethernet frame and structured as high-level network protocols. The high-level network protocol information carried inside the data field of each Ethernet frame is what actually establishes communications between applications running on computers attached to the network. The most widely used system of high-level network protocols is called the Transmission Control Protocol/Internet Protocol (TCP/IP) suite.
The important thing to understand is that the high-level protocols are independent of the Ethernet system. There are several network protocols in use today, any of which may send data between computers in the data field of an Ethernet frame. In essence, an Ethernet LAN with its hardware and Ethernet frame is simply a trucking service for data being sent by applications. The Ethernet LAN itself doesn't know or care about the high-level protocol data being carried in the data field of the Ethernet frame.
Since the Ethernet system is unaffected by the contents of the data field in the frame, different sets of computers running different high-level network protocols can share the same Ethernet. For example, you can have a single Ethernet that supports four computers, two of which communicate using TCP/IP, and two that use some other system of high-level protocols. All four computers can send Ethernet frames over the same Ethernet system without any problem.
The details of how network protocols function are an entirely separate subject from how the Ethernet system works and are outside the scope of this book. However, Ethernets are installed to make it possible for applications to communicate between computers using high-level network protocols to facilitate the communication. Let's take a quick look at one example of high-level network protocols to see how the Ethernet system and network protocols work together.
Network protocols are easy to understand since we all use some form of protocol in daily life. For instance, there's a certain protocol to writing a letter. We can compare the act of composing and delivering a letter to what a network protocol does to see how each works. The letter has a well-known form that has been "standardized" through custom. The letter includes a basic message with a greeting to the recipient and the name of the sender. After you're through writing the letter, you stuff it into an envelope, write the name and address of both the recipient and sender on the envelope, and give it to a delivery system, such as the post office, which handles the details of getting the message to the recipient's address.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Chapter 3: The Media Access Control Protocol
The tutorial in Chapter 2, introduced the Ethernet system and provided a brief look at how it works. In this chapter we take a much more detailed look at the original mode of operation used for Ethernet, which is based on the CSMA/CD media access control (MAC) protocol. This is also called half-duplex mode, to distinguish it from the full-duplex mode of operation. Full-duplex mode, which has no need for a MAC protocol, is described in Chapter 4.
In the original half-duplex mode, the MAC protocol allows a set of stations to compete for access to a shared Ethernet channel in a fair and equitable manner. The protocol's rules determine the behavior of Ethernet stations, including when they are allowed to transmit a frame onto a shared Ethernet channel and what to do when a collision occurs.
Since there is no central controller in an Ethernet system, each Ethernet interface operates independently while using the same MAC protocol. By equipping all interfaces with the same set of rules, all stations connected to a shared Ethernet channel operate the same way. Therefore, the MAC protocol functions as a kind of "Robert's Rules for Robots."
You don't need to know all the details of the MAC protocol in order to build and use Ethernet LANs. However, an understanding of the MAC protocol can certainly help when designing networks or troubleshooting problems. The media guidelines and MAC protocol are not some arbitrary laws dreamed up by a standards committee, but instead arise from the basic design and operation of the Ethernet system.
To simplify the description of the MAC protocol, this chapter has been broken down into three parts. The first part looks at the structure of the frame and the media access control rules used for transmitting frames, as well as how those rules affect the design of an Ethernet LAN. The next section takes a closer look at the operation of the important collision detect mechanism. Collision detection in Ethernet is frequently misunderstood, and we'll explain how it works, and how to design networks so that the collision detect system can do its job correctly. The last portion of the chapter describes how the high-level network software on your computer uses Ethernet frames to send data.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
The Ethernet Frame
The Ethernet specifications determine both the structure of a frame and when a station is allowed to send a frame. Ethernet media access control is based on Carrier Sense with Multiple Access and Collision Detect, which gives rise to the CSMA/CD acronym described in Chapter 2. The Ethernet system exists to move frames carrying application data between computers; the organization of the frame is central to the operation of the system.
The frame was first defined in the original Ethernet DEC-Intel-Xerox (DIX) standard, and was later redefined in the IEEE 802.3 standard, which is now the official Ethernet standard. The changes between the two standards were mostly cosmetic, except for the type field.
The DIX standard defined a type field in the frame. The first 802.3 standard (published in 1985) specified this field as a length field, with a mechanism that allowed both versions of frames to coexist on the same Ethernet. As it happens, most networking software kept using the type field version of the frame, and the IEEE 802.3 standard was recently changed to define this field as being either length or type, depending on usage.
Figure 3.1 shows the DIX and IEEE versions of the Ethernet frame. Since DIX and IEEE frames are identical in terms of the number and length of fields, Ethernet interfaces can be used to send either kind of frame. The only difference in the frames is in the contents of the fields and the subsequent interpretation of those contents by the stations that send and receive the frames. Next, we'll take a detailed tour of the frame fields. This tour will describe what each frame field does and what the differences are between the two versions of the frame.
Figure 3.1: DIX Ethernet and IEEE 802.3 frames
The frame begins with the 64-bit preamble field, which allows 10 Mbps Ethernet interfaces on the network to synchronize themselves with the incoming data stream before the important data fields arrive. The preamble exists in order to allow the beginning of the frame to lose a few bits due to signal start-up delays as it travels through a 10 Mbps system. This protects the rest of the frame from these effects. Like the heat shield of a spacecraft, which protects the spacecraft from burning up during re-entry, the preamble is the shield that protects the bits in the rest of the frame.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Media Access Control Rules
Now that we've seen what the structure of a frame is, let's look at the rules used for transmitting a frame on a half-duplex shared Ethernet channel. When transmitting a frame the station goes through the following steps:
  • When a signal is being transmitted on the channel, that condition is called carrier .
  • When a station attached to an Ethernet wants to transmit a frame, it waits until the channel goes idle, as indicated by an absence of carrier .
  • When the channel becomes idle, the station waits for a brief period called the interframe gap (IFG), and then transmits its frame.
  • If two stations happen to transmit simultaneously, they detect the collision of signals and reschedule their frame transmission. This occurrence is referred to as collision-detect.
There are two major things an interface connected to a half-duplex channel must do when it wants to send a frame. It must figure out when it can transmit, and it must be able to detect and respond to a collision. We'll first look at how the interface figures out when to transmit, and then we'll describe the collision detect mechanism.
The rules governing when an interface may transmit a frame are simple:
  1. If there is no carrier (i.e., the medium is idle) and the period of no carrier has continued for an amount of time that equals or exceeds the IFG, then transmit the frame immediately. If a station wishes to transmit multiple frames, it must wait for a period equal to the IFG between each frame.
    The IFG is provided to allow a very brief recovery time between frame reception for the Ethernet interfaces. IFG timing is set to 96 bit times. That is 9.6 microseconds (millionths of a second) for the 10 Mbps varieties of Ethernet, 960 nanoseconds (billionths of a second) for the 100 Mbps varieties of Ethernet, and 96 nanoseconds for Gigabit Ethernet.
  2. If there is a carrier (i.e., the channel is busy), then the station continues to listen until the carrier ceases (i.e., the channel is idle). This is known as deferring
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Essential Media System Timing
While signals travel very fast on an Ethernet, they still take a finite amount of time to propagate over the entire media system. The longer the cables used in the media system, the more time it takes for signals to travel from one end of the system to the other. The total round-trip time used in the slot time includes the time it takes for frame signals to go through all of the cable segments. It also includes the time it takes to go through all other devices, such as transceiver cables, transceivers and repeaters.
The maximum length for cable segments are carefully designed so that the essential signal timing of the system is preserved, even if you use maximum-length segments everywhere and the system is the largest allowed. The guidelines for each media variety incorporate the essential timing and round-trip signal delay requirements needed to make any half-duplex Ethernet up to the maximum-size system work properly. The cable segment guidelines are described in the media system chapters located in Part II. The correct signal timing is essential to the operation of the MAC protocol, so let's look at the slot time in more detail.
The total set of round-trip signal delays are summed up in the Ethernet slot time, which is defined as a combination of two elements:
  • The time it takes for a signal to travel from one end of a maximum-sized system to the other end and return. This is called the physical-layer round-trip propagation time .
  • The maximum time required by collision enforcement, which is the time required to detect a collision and to send the collision enforcement jam sequence.
Both elements are calculated in terms of the number of bit times required. Adding the two elements together plus a few extra bits for a fudge factor gives us a slot time that is 512 bit times for 10 and 100 Mbps systems. The Gigabit Ethernet slot time is described later in this chapter.
The time it takes to transmit a frame that is 512 bits long is slightly longer than the actual amount of time it takes for the signals to get to one end of a maximum size Ethernet and back. This includes the time required to transmit the jam sequence. Therefore, when transmitting the smallest legal frame, a transmitting station will always have enough time to get the news if a collision occurs, even if the colliding station is at the other end of a maximum-sized Ethernet.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Collision Detection and Backoff
Collision detection and backoff is an important feature of the Ethernet MAC protocol. It's also a widely misunderstood and misrepresented feature. Let's clear up a couple of points right away:
  • Collisions are not errors. Instead, collisions are a normal part of the operation of an Ethernet LAN. They are expected to happen, and collisions are handled quickly and automatically.
  • Collisions do not cause data corruption. As we've just seen, when a collision occurs on a properly designed and implemented Ethernet, it will happen sometime in the first 512 bit times of transmission. Any frame transmission that encounters a collision is automatically resent by the transmitting station. Any frame less than 512 bits long is considered a collision fragment, and is automatically and silently discarded by all interfaces.
As noted in the last chapter, it's unfortunate that the original Ethernet design used the word "collision" for this aspect of the Ethernet MAC protocol. Despite the name, collisions are not a problem on an Ethernet. Instead, the collision detect and backoff feature is a normal part of the operation of Ethernet, and results in fast and automatic rescheduling of transmissions.
The collisions counted by a typical Ethernet interface are the ones that occur while that interface is trying to transmit a frame. Collision rates can vary widely, depending on the traffic rate on the Ethernet channel and on the number of transmitting stations. Even on very lightly loaded networks, collisions will occur occasionally. Collision rates may range anywhere from fractions of a percent of the number of frame transmissions on up to a larger percentage on heavily loaded networks. Collision rates can be significantly higher for very heavily loaded networks, such as a network supporting high-speed computers.
In any case, the thing to worry about is the total traffic load on the network. Since the collision rate is simply a reflection of the normal functioning of an Ethernet, the rate of collisions seen by a given interface is not significant. Chapter 19, provides more information about measuring the performance of an Ethernet channel.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Gigabit Ethernet Half-Duplex Operation
The Gigabit Ethernet half-duplex mode uses the same basic CSMA/CD access mechanism as the 10 and 100 Mbps varieties of Ethernet, with the major exception of the slot time. The slot time in Gigabit Ethernet was modified to accommodate the special timing constraints which arise from the speed of the system.
Currently, all Gigabit Ethernet equipment is based on the full-duplex mode of operation described in Chapter 4. To date, none of the vendors have plans to develop equipment based on half-duplex Gigabit Ethernet operation. Nonetheless, a half-duplex CSMA/CD mode for Gigabit Ethernet has been specified, if only to insure that Gigabit Ethernet met the requirements for inclusion in the IEEE 802.3 CSMA/CD standard. For the sake of completeness, a description of Gigabit Ethernet half-duplex mode is provided here.
A major challenge for the engineers writing the Gigabit Ethernet standard was to provide a sufficiently large network diameter in half-duplex mode. As we've seen, the maximum network diameter (i.e., cable distance) between any two stations largely determines the slot time, which is an essential part of the CSMA/CD MAC mechanism.
Repeaters, transceivers and the interfaces have circuits that require some number of bit times to operate. The combined set of these devices used on a network takes a significant number of bit times to handle frames, respond to collisions, and so on. It also takes a small amount of time for a signal to travel over a length of fiber optic or metallic cable. All of this results in the total timing budget for signal propagation through a system, which determines the maximum cabling diameter allowed when building a half-duplex Ethernet system.
In Gigabit Ethernet, the signaling happens ten times faster than it does in Fast Ethernet, resulting in a bit time that is one-tenth the size of the bit time in Fast Ethernet. Without any changes in the timing budget, the maximum network diameter of a Gigabit Ethernet system would be about one-tenth of that for Fast Ethernet, or in the neighborhood of 20 meters (65.6 feet).
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Collision Domain
A useful concept to keep in mind while working with Ethernet is the notion of collision domain . This term refers to a single half-duplex Ethernet system whose elements (cables, repeaters, station interfaces and other network hardware) are all part of the same signal timing domain. In a single collision domain, if two or more devices transmit at the same time a collision will occur. A collision domain may encompass several segments, as long as they are linked together with repeaters, as shown in Figure 3.4.
Figure 3.4: Ethernet collision domain
A repeater is a signal-level device that enforces the collision domain on the segments to which it is connected. The repeater only concerns itself with individual Ethernet signals; it does not make any decisions based on the addresses of the frame. Instead, a repeater simply retransmits the signals that make up a frame.
Repeaters make sure that the repeated media segments are part of the same collision domain by enforcing any collisions seen on any segment attached to the repeater. For example, a collision on segment A is enforced by the repeater sending a jam sequence onto segment B. As far as MAC protocol (including the collision detection scheme) is concerned, a repeater makes multiple network cable segments function like a single cable. Repeaters are described in more detail in Chapter 17.
On a given Ethernet composed of multiple segments connected with repeaters, all of the stations are involved in the same collision domain. The collision algorithm is limited to 1024 distinct backoff times. Therefore, the maximum number of stations allowed in the standard for a multi-segment LAN linked with repeaters is 1024. However, that doesn't limit your site to 1024 stations, because Ethernets can be connected together with packet switching devices such as switching hubs or routers.
As Figure 3.5 illustrates, the repeaters and computers are connected by means of a switching hub . These Ethernets are in separate collision domains since switching hubs do not forward collision signals from one segment to another. Switching hubs contain multiple Ethernet interfaces. They operate by receiving a frame on one Ethernet port, moving the data through the hub, and transmitting the data out another Ethernet port in a new frame.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Ethernet Channel Capture
The Ethernet MAC protocol is a reliable, low overhead access control system that has proved its worth in millions of Ethernets. However, the MAC protocol is not perfect, and there are aspects of its operation that are not always optimal.
The best known example of this is an effect called Ethernet channel capture . Channel capture results in short-term unfairness, during which a station tends to consistently lose the competition for channel access. There must be one or more stations with a lot of data to send for channel capture to occur, causing them to contend for access the channel.
The sending station must also be able to continually transmit data at the maximum rate supported by the channel for short periods of time. Continuously sending data in this way sets up the collision and backoff conditions required for channel capture to occur. If the sending station does not have enough performance to continuously send data at the full rate supported by the channel, then the capture effect will not occur.
An example of channel capture works as follows. If you look at all the stations on a channel when several active stations are simultaneously contending for access, you would expect to see collisions. Each station will possess a nonzero collision counter. As soon as one of the stations acquires the channel and delivers its frame, it clears its collision counter and starts over with a new frame transmission. The rest of the stations trying to transmit will still have nonzero collision counters.
If the winning station immediately returns to the channel contention with a collision counter of zero, it has an advantage over the other stations which have higher collision counters. The stations with higher collision counters will tend to choose longer backoff times before retrying their frame transmissions. The station that wins will return to the fray with a zero collision counter, and therefore tend to continue winning. This station will effectively capture the channel for a brief period.
This can only occur if the winning station can rapidly and continually transmit data, which requires a high-performance station with a lot of data to send. This is not a common scenario, since most stations typically send data in short bursts. This helps explain why channel capture was first noticed when artificially high network loads were created using performance test software. Another place channel capture may be seen is when a file server is generating large bulk data flows doing backups while various user machines are trying to access the same channel.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
High-level Protocols and the Ethernet Frame
A wide range of computers can use the same Ethernet system, with each computer "speaking" several different high-level network protocols. The process of identifying which high-level network protocol suite is being carried in the data field of the frame is called multiplexing. Multiplexing simply means that multiple sources of information can be placed onto a single system. In this case, multiple high-level protocols can be carried over the same Ethernet system.
The original system of multiplexing for Ethernet is based on using the type field in the Ethernet frame. For example, the high-level protocol software can create a packet of IP data, and then hand the packet to the software on the computer that understands how to create Ethernet frames with type fields. The framing software inserts a hexadecimal value in the type field of the frame that corresponds to the type of high-level protocol being carried by the frame, and then hands the data to the interface driving software.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Chapter 4: Full-Duplex Ethernet
Full-duplex is an optional mode of operation allowing simultaneous communication between a pair of stations. The link between the stations must use a point-to-point media segment, such as twisted-pair or fiber optic media, to provide independent transmit and receive data paths. With full-duplex mode enabled, both stations can simultaneously transmit and receive, which doubles the aggregate capacity of the link. For example, a half-duplex Fast Ethernet twisted-pair segment provides a maximum of 100 Mbps of bandwidth. When operated in full-duplex, the same 100BASE-TX twisted-pair segment can provide a total bandwidth of 200 Mbps.
Another major advantage of full-duplex operation is that the maximum segment length is no longer limited by the timing requirements of shared channel half-duplex Ethernet. In full-duplex mode, the only limits are those set by the signal-carrying capabilities of the media segment. This is especially useful for fiber optic segments.
The optional full-duplex mode is specified in the 802.3x supplement to the standard, which formally describes the methods used for full-duplex operation. This supplement was approved for adoption into the IEEE 802.3 standard in March 1997. The 802.3x supplement also describes an optional set of mechanisms used for flow control over full-duplex links. The mechanisms used to establish flow control are called MAC Control and PAUSE. First we'll describe how full-duplex mode works, and then we'll show how the MAC Control and PAUSE mechanisms can be used to provide flow control over a full-duplex link.
The following requirements, as stated in the 802.3x standard, must be met for full-duplex operation:
  • The media system must have independent transmit and receive data paths that can operate simultaneously. Such data paths are typically found on twisted-pair and fiber optic links.
  • There are exactly two stations connected with a full-duplex point-to-point link. Since there is no contention for use of a shared medium, the multiple access algorithms (i.e., Carrier Sense with Multiple Access and Collision Detect, or CSMA/CD) are unnecessary.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Operation of Full-Duplex
The following requirements, as stated in the 802.3x standard, must be met for full-duplex operation:
  • The media system must have independent transmit and receive data paths that can operate simultaneously. Such data paths are typically found on twisted-pair and fiber optic links.
  • There are exactly two stations connected with a full-duplex point-to-point link. Since there is no contention for use of a shared medium, the multiple access algorithms (i.e., Carrier Sense with Multiple Access and Collision Detect, or CSMA/CD) are unnecessary.
  • Both stations on the LAN are capable of, and have been configured to use, the full-duplex mode of operation. This means that both Ethernet interfaces must have the capability to simultaneously transmit and receive frames.
The original mode of Ethernet operation is half-duplex, based on CSMA/CD. The operation of CSMA/CD is described in detail in Chapter 3. In the CSMA/CD-based half-duplex mode of operation, only one station can transmit at any given time; other stations must defer until that transmission is complete.
Ethernet repeaters as defined in the IEEE 802.3 standard are not stations and cannot be operated in full-duplex mode. Ethernet repeaters simply pass Ethernet signals between attached segments, and a repeater can only operate in half-duplex mode.
Beware of confusing terms that you may find in advertising or network trade journals, such as "full-duplex repeater." Vendors have created devices which they call "buffered repeaters," "buffered distributors," and "full-duplex repeaters," but these are not true repeaters. The operation and configuration of these devices is not described in the Ethernet standard and system configuration guidelines. Instead, these devices typically operate as switching hubs. Switching hubs are described in Chapter 18.
Figure 4.1: Full-duplex operation
Figure 4.1 shows two stations simultaneously sending and receiving over a full-duplex link segment. The segment provides independent data paths so that both stations can be active without interfering with one another's transmissions. To provide full-duplex operation, both the station interface (
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Ethernet Flow Control
The rate of traffic on network backbones is always growing, and as a result backbone switches connected together with full-duplex links can be heavily loaded with traffic. A switching hub typically has a fixed set of resources, in the form of internal switching bandwidth and packet buffers, which it apportions to its switching ports. Resources like packet buffer memory are expensive, and many low-cost switches limit these resources. To keep these limited resources from being overwhelmed, a variety of non-standard flow control mechanisms were developed by switching hub vendors for use on half-duplex segments. These include the use of a short burst of carrier signal sent by the switching hub to cause stations on a half-duplex segment to stop sending data when the buffers on a switching port are full. These mechanisms are described in more detail in Chapter 18.
This sort of mechanism is based on the half-duplex mode of operation—it will not work on a full-duplex segment that is not using the CSMA/CD algorithm and ignores things like carrier. As a result, a switching hub connected to full-duplex segments needs a new mechanism to send a flow control message. To that end, an explicit flow control message is provided by the optional MAC Control and PAUSE specifications in the 802.3x Full-Duplex supplement.
The optional MAC Control portion of the 802.3x supplement provides a mechanism for real-time control and manipulation of the frame transmission and reception process in an Ethernet station. In normal Ethernet operation, the media access control (MAC) protocol defines how to go about transmitting and receiving frames. In the optional Ethernet flow control system, the MAC Control protocol provides mechanisms to control when Ethernet frames are sent.
When implemented, the MAC Control system provides a way for the station to receive a MAC Control frame and act upon it. The operation of the MAC Control system is transparent to the normal media access control functions in a station. MAC Control is not used for a non-real-time function like configuring interfaces, which is handled by network management mechanisms. Instead, MAC Control is designed to allow stations to interact in real time to control the flow of traffic. New functions beyond flow control may be added in the future.
Additional content appearing in this section has been removed.
Purchase this book now or read it online at Safari to get the whole thing!
Chapter 5: Auto-Negotiation
Automatic configuration of Ethernet equipment is provided by the Auto-Negotiation protocol, which is defined in the Ethernet standard. This chapter describes the Auto-Negotiation protocol, and shows how automatic configuration functions. The need for an automatic configuration system becomes obvious when you consider the challenge facing someone who is installing a desktop computer and connecting it to an Ethernet system.
Among the things the installer needs to know are which speed should be set on the Ethernet interface and whether full-duplex mode should be enabled. However, these features are embedded in the network equipment and are invisible to the installer. One twisted-pair port looks a lot like another, and it is not obvious which network options may be supported. The Auto-Negotiation protocol allows Ethernet equipment to automatically select the correct speed and other features, thus relieving the installer of this configuration task.
The specifications for Auto-Negotiation were first published in 1995 as part of the 802.3u Fast Ethernet supplement to the IEEE standard. These specifications were based on an automatic configuration system called NWay, which was invented by National Semiconductor. Engineers working on the standard found that an automatic configuration signaling system could be readily developed that would work on twisted-pair links. Therefore, all Ethernet media systems that use twisted-pair media can also support the Auto-Negotiation signals. However, fiber optic links use a variety of light sources and optical wavelengths, which do not interoperate. That, in turn, makes it impossible to develop an automatic configuration signaling system that works on all fiber optic links.
For that reason, there is no IEEE standard Auto-Negotiation support for most fiber optic link segments. The only exception is the Gigabit Ethernet fiber optic automatic configu