ITNW 2313
LAN Hardware/Wiring & Installation

Lesson 5 - Exploring Access Methods

Instructor: Kenneth D. Stewart

Key Terms
access method channel band width access control polling front-end processor over saturated predictable access contention garbling	collision carrier sensing CSMA collision detection CSMA/CD collision avoidance CSMA/CA time slice probabilistic access token passing	token authorizing packet non-contention deterministic access performance characteristics network load network throughput perceived performance logical bus access methods logical ring access methods
All images are "clickable" links to larger, clearer images.

Access Methods

Access methods, sometimes called channel access methods, are theoretically independent of the topologies you just learned about. In reality, however, only a few combinations of physical and logical topologies work well together.

When several entities share the same communications medium (channel), some mechanism must be in place to control access fairly. It is unproductive to have everyone in a meeting speak at once, so rules of order were defined long ago for managing meetings. Similar rules, or access methods, are applied to networks.

In this lesson, you will learn about the three most common channel access methods and the topologies with which they are associated. These access methods are as follows:

• Polling
• Contention
• Token passing

The channel access methods discussed in the following sections include general rules that govern the devices as they access and transmit across the channel. Access methods use a certain amount of the channel's band width for access control. The usable portion of the channel's band width is limited by the access method being used. Each method has a different effect on network traffic.

Polling

Polling resembles a well-ordered meeting in which the chairman must recognize an attendee before that person is allowed to speak. The chairman's responsibility is to maintain order in the meeting and ensure that each person who wants to speak has an opportunity to do so. Polling is most closely associated with mainframe (point-to-point) computer networks. By using polling, one device such as a mainframe front-end processor, is designated as the primary device. Primaries also are known as the channel access administrators, controllers, or masters. All access to the network is controlled by the primary.

The primary queries (polls) each of the secondary devices, also known as slaves. As each secondary is polled, the primary inquires if the secondary has information to be transmitted. Only when it is polled does the secondary have access to the communication channel. Each system has rules pertaining to how long each secondary can transmit data. The process of polling is much like a committee chairman who asks each member in turn to vote on an issue. Polling can be utilized in virtually any network topology.

The advantages of polling are as follows:

• Many characteristics of polling can be determined centrally, including the polling order and node priorities.

• Polling ensures that channel access is predictable and fixed. Because the time delays between the primary and secondary devices can be calculated, this access method is called deterministic. Deterministic access methods are suitable for controlling some automated equipment because each piece of equipment is guaranteed access to the network at predetermined intervals.

• Polled channels cannot be over saturated with traffic. As demand increases, traffic increases up to a maximum level. The polling mechanism ensures that maximum traffic level cannot be exceeded. Nor can excess traffic reduce the performance of the network.

Polling has the following disadvantages:

• Some applications cannot function with the time delays required for polling other devices.

• The process of polling involves large numbers of messages that take up available band width. Traffic is required to poll each node, even nodes that are idle.

• Some polled networks use half-duplex transmission lines. This means that the primary and secondary devices must "turn around" the line, requiring some band width.

• Polling requires a sophisticated central control mechanism that requires extensive configuration.

Contention

Contention on a network resembles conversation in a meeting. Every attendee can attempt to speak at any time. When two speakers interfere, however, the conversation is garbled, and the speakers must begin again. Any speaker can speak at any time, and the speakers must contend for openings in the conversation in which to state their messages.

Similarly, on a contention network, any device can transmit whenever it needs to send information. To avoid data collisions, specific contention protocols were developed requiring the device to listen to the cable before transmitting information.

Contention is also known as random access because, unlike polling and token passing, there is no fixed order in which the nodes can transmit.

STUDY NOTE: The act of "listening" to the channel to see if any traffic exists is called carrier sensing, and contention-based networks are called Carrier Sense Multiple Access (CSMA) networks.

Even though each station listens for network traffic before it attempts to transmit, it remains possible for two transmissions to overlap on the network. This overlap is called a collision. As the diagram shows, collisions occur because it takes time for signals to propagate through the network. Both stations A and D have found the network clear and transmit a message. A few micro-seconds are required for the signal from A to reach D. During that period, D is free to transmit, and a collision can occur.

As a result of collisions, access to a CSMA network is somewhat unpredictable, and CSMA networks can be referred to as random or statistical access networks.

Collisions are part of the normal operation of a CSMA network. Two specialized methods of collision management have been developed to improve performance: Collision Detection (CD) and Collision Avoidance (CA).

The collision detection approach listens to the network traffic as the card is transmitting. By analyzing network traffic, it is possible to detect collisions and initiate retransmissions. Carrier Sense Multiple Access with Collision Detection (CSMA/CD) is the access method utilized in Ethernet and IEEE 802.3.

Collision avoidance uses time slices to make network access smarter and avoid collisions. Carrier Sense Multiple Access with collision avoidance is the access mechanism used in Apple's LocalTalk network.

Contention offers the following benefits:

• Contention is a very simple access method that has low administrative overhead requirements. No network traffic is necessary to manage the access scheme.

• Actual user data throughput is rather high at low traffic levels in comparison to the total amount of utilized network band width.

The disadvantages of contention are as follows:

• At high traffic levels, data collisions and the resulting retransmission diminish performance dramatically. It is theoretically possible that collisions can be so frequent at higher traffic levels that no station has a clear chance to transmit.

• Channel access is probabilistic  rather than deterministic. Because of retransmissions and the time it takes to sense collisions, automated equipment that cannot tolerate delays cannot use this type of access. Contention offers no means of establishing the frequency of a station's opportunities to transmit.

Examples of networks that use contention are as follows:

• Ethernet (original)
• IEEE 802.3 & IEEE 802.3u
  (10Base5, 10Base2, 10Base-T, & 100Base-TX)
• LocalTalk (Apple)

Token Passing

Token passing  resembles a children's story-telling game in which the players pass a ball around a circle. When a player receives the ball, he or she is expected to tell part of a story. Players can talk only when the ball is in their possession.

Token passing uses a special authorizing packet of information to inform devices that they can transmit data. These packets are called tokens and are passed around the network in an orderly fashion from one device to the next. Devices can transmit only if they have control of the token. This method distributes the access control among all the devices.

Two approaches to token passing are available. Token Ring uses a ring topology. Each station passes the token to the next station in the ring. ARCnet also uses token passing; however, with ARCnet, each station passes the token to the station with the next higher node address, regardless of its physical location on the network (token passing bus).

The diagram below shows examples of token passing in Token Ring and ARCnet networks.

Advantages of Token Passing

Token passing provides the following advantages:

• Token passing offers the highest data throughput possible under high traffic conditions. Only one transmission can occur at a time, and collisions cannot occur (non-contention). Therefore, token passing experiences less performance degradation at higher traffic levels than contention.

• Token passing is deterministic. Each station is guaranteed an opportunity to transmit each time the token travels around the ring.

• Some token passing systems enable you to set priorities for devices that need controlled access to the token.

• As the traffic increases, data throughput also increases to a certain level, and then stabilizes.

The disadvantages of token passing are as follows:

• Token passing involves complicated protocols for managing the network and recovering from errors. The traffic associated with these protocols has higher band width overhead then is required for CSMA.

• All devices require complicated software that needs to be modified whenever a station is added or removed.

• Some systems require an additional central controller that adds to the overhead and reduces throughput. Cabling and network hardware can be more expensive for token passing networks than for CSMA networks.

Examples of token passing networks include the following:

• IEEE 802.4, also known as token bus. Token bus uses token passing access control and a bus topology. (similar to ARCnet)

• IEEE 802.5, also known as Token Ring . Token Ring uses token passing access control and a star-wired-ring topology.

• ARCnet  uses token passing based on node addresses, using a star-wired network with a logical bus topology.

• TokenTalk is Apple's standard for networking Macintosh computers on Token Ring networks.

Performance Comparisons

Token passing and CSMA, the most common access methods used in LANs, have different performance characteristics (see the graph). The "Load"  x-axis represents the demand being placed on the network. The " Throughput"  y-axis represents the data actually being transmitted.

Notice that the throughput of a CSMA  network rises smoothly with increased traffic levels up to a point. At that point, collisions begin to occur with greater frequency, resulting in a gradual reduction in network throughput. At some point, network throughput reaches unacceptably low levels.

Token passing  exhibits reduced performance at lower traffic levels than CSMA. This is a result of the many administrative mechanisms required for token access. Throughput rises smoothly until the network is fully utilized. At that point, throughput stabilizes. Throughput does not degrade because no collisions can occur. However, beyond the plateau, all workstations are sharing a strictly limited band width. Although total throughput remains stable, the bandwidth available to a given station diminishes as demand increases.

The user's perception  is that the network's performance is diminishing as the load demand increases. This graph above illustrates how throughput decreases as a percentage of demand. Basically, as demand increases, a smaller percentage of the demand can be satisfied. With contention-based networks, the fall-off after a certain point is fairly rapid until the number of collisions interferes with virtually all traffic on the network and few, if any, packets are actually delivered.

Perceived performance of a token passing network also declines, but never reaches zero. Each user is guaranteed a fair share of the network's band width (deterministic), although this share may, at some point, be considered inadequate for the user's needs.

Combining Architectures and Access Methods

The following table summarizes common types of networks in terms of their topologies and access methods.

Notice that CSMA technologies are only applicable in networks that are logical buses. This is the case, because each station must be able to sense all network traffic to determine if the network is busy. On a ring network, a node can detect only the data transmissions that happen to pass it, and cannot determine whether traffic exists elsewhere on the network. CSMA is, therefore, not applicable to rings.

On the other hand, the only access method that works with logical rings  is token passing. This is the case because each station must be able to receive each packet it transmits after that packet travels around the logical ring.

Prior Lesson

Index

Next Lesson

  This page is maintained by:   Kenneth D. Stewart

---

	viking.delmar.edu Last modified 08 Jan 2002		kstewart@delmar.edu Copyright © 2002