Basic Data Communications Series
Email Home	2.0 Network Protocols 2.1 Introduction As already stated the top four layers of the OSI reference archetecture are normally resident in the host computer or other intelligent system that requires to communicate via the network to another host or system. The bottom three layers are normally resident in the communication equipment that controls the access and transmission/ reception of the data. The Potocols are the rules (algorithms) which goveren the orderly connection and control of the data communication onto the network media. The protocols reside in layers 2 and 3 of the media access control equipement (DCE) and this is where the real differences between the various types of network are implimented. Thus various standard protocls are implimented depending on the network ie LAN WAN, packet switched, message switched etc. The system model of a network node is quite often represented as above. Here the upper four layers of the OSI model resident in the host computer are designated Data Terminal Equipment (DTE) and the lower three layers are resident in the network communication equipment or Data Circuit Terminating Equipment (DCE) The DCE is further decomposed into the Media Access Control (MAC) Layer and the Physical Layer. The MAC layer contains layers 2 and 3 of the OSI model and the PHY layer contains layer 1 of the OSI model. We will now consider how the various MAC layer protocols may be classified. 2.2 Protocol Classification The following discussion will be limited to non-centralised protocols i.e. the same set of rules are implemented at all nodes with no permanent overall controller. Two main types of multiple-access protocols can be defined: conflict-free and contention. The former rely on the channel being orderly allocated to nodes in turn. Contention schemes differ in that the channel is a ‘free-for-all’ and contentions when they occur must be resolved by the protocol. The resolution rules can be static or dynamic, the former relying on a set of fixed rules regardless of the system dynamics (traffic density, interference etc), the latter taking advantage of and tracking the system changes. Taxonomy of multiple access protocols [1] is illustrated in Fig2-1. The practical protocols that are classified in this way form a very large set; therefore the following brief survey will concentrate on some of the more common types and give a brief description of each. 2.3 Conflict Free Protocols 2.3.1 Frequency Division Multiple Access FDMA divides the available frequency band into a number of discrete sub-frequencies, each one allocated to a user. The advantages of FDMA are simplicity, however when a user is idle a portion of the available bandwidth is lost. Also FDMA is not flexible in that an additional node will have impacts on the complete system. FDMA is an example of a static allocation conflict free protocol. 2.3.2 Time Division Multiple Access TDMA divides the channel access into a number of time slots repeated cyclically as a frame. Each time slot is uniquely allocated to a user. A disadvantage of TDMA is that the time slot length bound messages; also bandwidth is lost when a user is idle. TDMA is used heavily in aerospace applications and is an example of a static allocation conflict free protocol. 2.3.3 Mini-Slotted Alternating Priority MSAP is similar to TDMA except that the time slots are not statically allocated thus overcoming the loss in bandwidth due to idle nodes. The slots are allocated dynamically on a priority basis and various schemes are possible. Allocated priorities may be fixed, round -robin or alternating. A fixed priority scheme assigns fixed priorities to user i, whereas a round-robin scheme shifts the priority amongst users cyclically. The alternating priority structure is similar to round robin except a node is allowed to transmit all queued messages before priority is re-allocated. MSAP is an example of a dynamic allocation conflict free protocol. 2.3.4 Token Ring A token ring network consists of a point-to-point ring of nodes in which a token message is passed from each node to its neighbour. A sending node stops the token circulation and transmits its data then passes the token as before. Additionally nodes may be prioritised dynamically by including a priority field in the token. Disadvantages include the additional overhead to ensure maintenance of the token passing protocol. Token ring is an example of a dynamic allocation conflict free protocol. 2.4 Contention Protocols 2.4.1 Aloha The oldest contention multiple access protocol is aloha in which nodes are allowed to transmit at will and detect collisions on the channel by concurrently receiving the transmitted message. When collisions occur they are randomly rescheduled to be re-transmitted at some time in the future. The advantage of aloha is its simplicity however its performance is poor due to the number of collisions and retransmissions that are necessary. The performance may be improved by time-slotting the access. Aloha is an example of a static resolution contention protocol. 2.4.2 Carrier Sense Protocols Carrier Sense protocols are a large class of Multiple Access schemes in which nodes are allowed to transmit at will but only if the channel is sensed and found idle. Collisions will still occur but not to the same extent as pure aloha. The variations in the CSMA protocols are due to the way the collisions are handled. Non-persistent CSMA In NP-CSMA a node sensing the channel busy behaves as if a collision had occurred and randomly re-schedules the retransmission at a later time. This protocol is thus an example of a static resolution contention protocol. One persistent CSMA With NP-CSMA there are times when the channel is idle due to the static re-scheduling of the re-transmission. In 1P-CSMA the node on sensing the channel busy persists in sensing the channel and transmits as soon as the channel becomes free. This protocol is thus an example of a dynamic resolution contention protocol. Slotted CSMA In slotted CSMA time slots are defined less than a packet length – mini slots. A node is constrained to sense the channel at the beginning of a mini-slot. If the channel is busy then either the NP-CSMA protocol or 1P-CSMA protocols is applied as above. Collision Detect CSMA In CSMA/CD the overhead caused by collisions is reduced by terminating an unsuccessful transmission (due to collision) as soon as possible. This is done by detecting the collision as soon as it happens and then randomly re-scheduling the transmission to a later time. Another feature is that of consensus reinforcement: after a collision all nodes generate a jam signal for a period t_cr in order to ensure that all network nodes confirm that a collision has taken place. This protocol is thus an example of a dynamic resolution contention protocol. Collision Avoidance CSMA CSMA/CA is similar to CSMA/CD. After consensus reinforcement each node starts contention time slots, assigned to a particular node. Each node is allowed to transmit during its contention time slot. The slots are rotated after each transmission in order to ensure fairness and determinacy; the inclusion of p-slots for global prioritisation is also an option. Two distinct versions of CSMA/CA protocols exist. If the number of slots equals the number of nodes, this is referred to as reservation CSMA/CA. The other variation has fewer slots than nodes and these are assigned randomly to minimise collisions. This protocol is an example of a dynamic resolution contention protocol. Non-Destructive CSMA This is also known as binary countdown or bit dominance CSMA. In CSMA/ND all nodes wait for an idle channel by carrier sensing before transmitting a message. Competing nodes resolve contention by unique node identification (or priority). The transmission medium has the characteristics that a logical assertion dominates or overrides the opposite assertion. Thus a logical false could for instance override a logical true. During the contention competition the node with a less dominant logical identification (or priority) looses and retries with possibly an enhanced priority. This protocol is an example of a static resolution contention protocol when dominance is set by node identification but can also be considered a dynamic resolution protocol if system transmission rules are included to adjust the priority of a dominated node. The above survey represents just a small selection of those available. Some other protocols are: Stop and Wait Go back N High Level data Link Control - HDLC X.25 TCP/IP

Basic Data Communications Series

2.0 Network Protocols

2.1 Introduction

As already stated the top four layers of the OSI reference archetecture are normally resident in the host computer or other intelligent system that requires to communicate via the network to another host or system.

The bottom three layers are normally resident in the communication equipment that controls the access and transmission/ reception of the data.

The Potocols are the rules (algorithms) which goveren the orderly connection and control of the data communication onto the network media. The protocols reside in layers 2 and 3 of the media access control equipement (DCE) and this is where the real differences between the various types of network are implimented. Thus various standard protocls are implimented depending on the network ie LAN WAN, packet switched, message switched etc.

The system model of a network node is quite often represented as above. Here the upper four layers of the OSI model resident in the host computer are designated Data Terminal Equipment (DTE) and the lower three layers are resident in the network communication equipment or Data Circuit Terminating Equipment (DCE)

The DCE is further decomposed into the Media Access Control (MAC) Layer and the Physical Layer. The MAC layer contains layers 2 and 3 of the OSI model and the PHY layer contains layer 1 of the OSI model.

We will now consider how the various MAC layer protocols may be classified.

2.2 Protocol Classification

The following discussion will be limited to non-centralised protocols i.e. the same set of rules are implemented at all nodes with no permanent overall controller. Two main types of multiple-access protocols can be defined: conflict-free and contention. The former rely on the channel being orderly allocated to nodes in turn.

Contention schemes differ in that the channel is a ‘free-for-all’ and contentions when they occur must be resolved by the protocol. The resolution rules can be static or dynamic, the former relying on a set of fixed rules regardless of the system dynamics (traffic density, interference etc), the latter taking advantage of and tracking the system changes.

Taxonomy of multiple access protocols [1] is illustrated in Fig2-1.

The practical protocols that are classified in this way form a very large set; therefore the following brief survey will concentrate on some of the more common types and give a brief description of each.

2.3 Conflict Free Protocols

2.3.1 Frequency Division Multiple Access

FDMA divides the available frequency band into a number of discrete sub-frequencies, each one allocated to a user. The advantages of FDMA are simplicity, however when a user is idle a portion of the available bandwidth is lost. Also FDMA is not flexible in that an additional node will have impacts on the complete system. FDMA is an example of a static allocation conflict free protocol.

2.3.2 Time Division Multiple Access

TDMA divides the channel access into a number of time slots repeated cyclically as a frame. Each time slot is uniquely allocated to a user. A disadvantage of TDMA is that the time slot length bound messages; also bandwidth is lost when a user is idle. TDMA is used heavily in aerospace applications and is an example of a static allocation conflict free protocol.

2.3.3 Mini-Slotted Alternating Priority

MSAP is similar to TDMA except that the time slots are not statically allocated thus overcoming the loss in bandwidth due to idle nodes. The slots are allocated dynamically on a priority basis and various schemes are possible. Allocated priorities may be fixed, round -robin or alternating. A fixed priority scheme assigns fixed priorities to user i, whereas a round-robin scheme shifts the priority amongst users cyclically. The alternating priority structure is similar to round robin except a node is allowed to transmit all queued messages before priority is re-allocated. MSAP is an example of a dynamic allocation conflict free protocol.

2.3.4 Token Ring

A token ring network consists of a point-to-point ring of nodes in which a token message is passed from each node to its neighbour. A sending node stops the token circulation and transmits its data then passes the token as before. Additionally nodes may be prioritised dynamically by including a priority field in the token. Disadvantages include the additional overhead to ensure maintenance of the token passing protocol. Token ring is an example of a dynamic allocation conflict free protocol.

2.4 Contention Protocols

2.4.1 Aloha

The oldest contention multiple access protocol is aloha in which nodes are allowed to transmit at will and detect collisions on the channel by concurrently receiving the transmitted message. When collisions occur they are randomly rescheduled to be re-transmitted at some time in the future. The advantage of aloha is its simplicity however its performance is poor due to the number of collisions and retransmissions that are necessary. The performance may be improved by time-slotting the access. Aloha is an example of a static resolution contention protocol.

2.4.2 Carrier Sense Protocols

Carrier Sense protocols are a large class of Multiple Access schemes in which nodes are allowed to transmit at will but only if the channel is sensed and found idle. Collisions will still occur but not to the same extent as pure aloha. The variations in the CSMA protocols are due to the way the collisions are handled.

Non-persistent CSMA

In NP-CSMA a node sensing the channel busy behaves as if a collision had occurred and randomly re-schedules the retransmission at a later time. This protocol is thus an example of a static resolution contention protocol.

One persistent CSMA

With NP-CSMA there are times when the channel is idle due to the static re-scheduling of the re-transmission. In 1P-CSMA the node on sensing the channel busy persists in sensing the channel and transmits as soon as the channel becomes free. This protocol is thus an example of a dynamic resolution contention protocol.

Slotted CSMA

In slotted CSMA time slots are defined less than a packet length – mini slots. A node is constrained to sense the channel at the beginning of a mini-slot. If the channel is busy then either the NP-CSMA protocol or 1P-CSMA protocols is applied as above.

Collision Detect CSMA

In CSMA/CD the overhead caused by collisions is reduced by terminating an unsuccessful transmission (due to collision) as soon as possible. This is done by detecting the collision as soon as it happens and then randomly re-scheduling the transmission to a later time. Another feature is that of consensus reinforcement: after a collision all nodes generate a jam signal for a period t_cr in order to ensure that all network nodes confirm that a collision has taken place. This protocol is thus an example of a dynamic resolution contention protocol.

Collision Avoidance CSMA

CSMA/CA is similar to CSMA/CD. After consensus reinforcement each node starts contention time slots, assigned to a particular node. Each node is allowed to transmit during its contention time slot. The slots are rotated after each transmission in order to ensure fairness and determinacy; the inclusion of p-slots for global prioritisation is also an option. Two distinct versions of CSMA/CA protocols exist. If the number of slots equals the number of nodes, this is referred to as reservation CSMA/CA. The other variation has fewer slots than nodes and these are assigned randomly to minimise collisions. This protocol is an example of a dynamic resolution contention protocol.

Non-Destructive CSMA

This is also known as binary countdown or bit dominance CSMA. In CSMA/ND all nodes wait for an idle channel by carrier sensing before transmitting a message. Competing nodes resolve contention by unique node identification (or priority). The transmission medium has the characteristics that a logical assertion dominates or overrides the opposite assertion. Thus a logical false could for instance override a logical true. During the contention competition the node with a less dominant logical identification (or priority) looses and retries with possibly an enhanced priority. This protocol is an example of a static resolution contention protocol when dominance is set by node identification but can also be considered a dynamic resolution protocol if system transmission rules are included to adjust the priority of a dominated node.

The above survey represents just a small selection of those available.

Some other protocols are:

Stop and Wait

Go back N

High Level data Link Control - HDLC

X.25

TCP/IP

Last changed: 04/11/2004, 15:28:11