Data Link Layer Issues - PowerPoint PPT Presentation

1 / 129
About This Presentation
Title:

Data Link Layer Issues

Description:

Frequency or time-division multiplexing can be used to share links in such a network ... Twisted pair is called a logical star topology, while still a physical ... – PowerPoint PPT presentation

Number of Views:74
Avg rating:3.0/5.0
Slides: 130
Provided by: randyf3
Category:
Tags: data | issues | layer | link | links | twisted

less

Transcript and Presenter's Notes

Title: Data Link Layer Issues


1
Data Link Layer Issues
  • Dealing with Different Types of Networks

2
Types of Networks
  • Network hardware can be categorized into
  • Circuit-switched (e.g. telephone)
  • Prior to communication, the hardware establishes
    a dedicated end-to-end connection
  • Since there is a dedicated connection, a
    continuous stream of bytes can be sent
  • Frequency or time-division multiplexing can be
    used to share links in such a network
  • Packet-switched (e.g. Ethernet, ATM)
  • Data is divided into packets of limited size, and
    each is forwarded through the network to the
    destination
  • This can be done by routers or switches

3
Types of Networks
  • Disadvantages
  • Circuit-switched
  • A dedicated connection that has no transmission
    means wasted bandwidth
  • A connection is time consuming if short,
    infrequent, or sporadic communication is to occur
  • Packet-switched
  • Forwarding each packet means that each router
    must decide the next hop for every packet (even
    for the same destination)
  • Routers are typically network slowdowns due to
    the amount of processing, as well as input/output
    buffering

4
Types of Networks
  • Circuit-switching is used in a telephone
    conversation
  • A connection to the receiver is established by
    the sender (the caller)
  • The telephone company reserves a certain
    bandwidth (64 Kbps for voice communication) for
    this call
  • If the bandwidth is not used by the callers, it
    is wasted
  • Packet-switching is similar to the postal service
  • Each message (envelope) is addressed to the
    recipient individually, and the postal service
    delivers each message to the recipient
  • The postal service may deliver these envelopes
    through different cities and methods of transport
    (airplane, truck, )
  • It can be said that these messages can be
    delivered using different routes

5
Circuit-Switching
A
B
Telephone Company Switching System
6
Circuit-Switching
A
B
Telephone Company Switching System
7
Packet-Switching
Quebec, QC
Buffalo, NY
Montreal, QC
A
B
Windsor, ON
Toronto, ON
Niagara Falls, ON
London, ON
Ottawa, ON
Kitchener, ON
Postal Network
8
Types of Packet-Switching
  • Virtual circuit-switching
  • A virtual circuit is created between source and
    destination
  • This VC is used for all subsequent sending of
    packets
  • Datagram
  • Each packet is routed individually

9
Virtual Circuit Packet-Switching
  • Advantages
  • After the first message, routing is faster
  • A route must only be determined once, for the
    first message
  • Once the route has been determined, the path used
    by the router is reused for all messages
  • As a result, routing tables are much smaller (and
    can be searched more quickly)
  • Because a connection is created, the connection
    identifier can be used (alone) to address packets
  • Typically, such as with ATM cells, this can
    reduce the size of a cell/packets header
  • Messages do not arrive out of order
  • As a result, receivers do not need to reorder the
    cells

10
Virtual Circuit Packet-Switching
  • Disadvantages
  • Connections take some time to create
  • Routers/switches must intercommunicate in order
    to create the connection
  • Infrequent messaging is not suitable for
    connection-based messaging
  • The connection may be lost after a timeout, and
    will have to be recreated again and again
  • The time delay for creating the connection may
    outweigh the speed benefits of using
    connection-based transport
  • Routing tables will be dynamic, and routing
    algorithms are more complex

11
Datagram Packet-Switching
  • Advantages
  • Connections need not be created
  • Infrequent messaging is perfect for
    connectionless messaging
  • Connectionless messaging can be resumed after any
    amount of delay, any number of times, without any
    delays due to the resumption of communication
  • Routing each message separately allows for load
    balancing
  • Some messages may be sent through one route, but
    when that route becomes saturated, messages may
    then be sent through a different route in order
    to achieve the most optimal communication possible

12
Datagram Packet-Switching
  • Disadvantages
  • Each message takes a certain amount of time to
    transmit (including transmission, routing,
    reception, etc.)
  • Nodes communicating large amounts of information
    in a short time will
  • Use a lot of bandwidth for things such as header
    information
  • Waste a lot of time routing messages to the same
    destination
  • Messages may arrive out of order
  • Messages must be reordered by the recipient

13
Multiple Access Strategies
  • Schemes for Sharing a Communication Medium

14
Multiple Access
  • Most networks are shared medium
  • This means that a single medium (e.g. radio
    frequency) is shared by all of a networks hosts
  • We need a scheme to allow the hosts to share the
    medium, without collisions
  • Collisions occur when two (or more) messages are
    transmitted at the same time
  • The result is constructive and destructive
    interference in the carrier wave
  • This causes the messages to be combined and
    scrambled

15
Contention
  • In contention networks, any node that has a
    packet to send, merely sends the packet
  • It is clear that this type of network frequently
    experiences collisions
  • The more nodes trying to communicate, the higher
    the chance of collisions
  • Thus, contention networks are severely limited in
    the number of hosts possible

16
Contention
Transmit
17
Contention
18
Contention
Transmit
19
Contention
20
Contention Collisions
Transmit
Transmit
21
Contention Collisions
Scrambled Signal
22
Contention
  • No collision avoidance is present
  • Messages are just sent
  • When collisions occur, the messages are simply
    resent after some random (or pseudo-random)
    amount of time
  • Collisions can occur anytime

23
Carrier Sensing
Test the medium for a signal
24
Carrier Sensing
Test the medium for a signal Available Transmit
25
Carrier Sensing
Test the medium for a signal
26
Carrier Sensing
Test the medium for a signal In use
27
Carrier Sensing
Transmission Complete
28
Carrier Sensing
Test the medium for a signal
29
Carrier Sensing
Test the medium for a signal Available
30
Carrier Sensing Collisions
31
Carrier Sensing Collisions
32
Carrier Sensing Collisions
Scrambled data
33
Carrier Sensing Collisions
Transmit
Transmit
34
Carrier Sensing (CSMA)
  • To reduce the number of collisions, the medium is
    tested for a signal before each transmission
  • If a signal exists, the node waits
  • Signal testing can be anything from detection of
    an electrical signal, to testing for photons
  • Collisions can still occur (although less often)
  • If a node tests for a signal before a
    transmission from another node, and transmits
    after, a collision occurs

35
Carrier Sensing Hardware
Transmitter
Receiver
If the message is broadcast or the address is
this stations address, the message is forwarded
to the receiver
When a signal is detected, transmissions
are blocked by the signal detector
Signal Detector
36
CSMA/CA
  • CSMA/CA networks (such as wireless 802.11g) also
    use carrier sensing and collision detect
  • However, detecting collisions in wireless
    networks is significantly more complicated
  • Also, after detecting carrier and determining
    there is no signal, a CSMA/CA network transmits a
    Do not broadcast message
  • If this message is sent without a collision, the
    host can assume it is safe to transmit

37
Carrier Sensing Networks
  • Advantages
  • No tokens
  • Simple hardware
  • No need for token transmission
  • Disadvantages
  • Collisions
  • Wasted bandwidth for re-transmits
  • Require complicated re-collision avoidance schemes

38
Token Passing
Transmit
Transmit
Transmit
Transmit
T
39
Token Passing
Transfer Token
T
40
Token Passing
T
41
Token Passing
Transmit
T
42
Token Passing
  • A small packet (the token) is passed from node to
    node
  • When a node has the token, it has sole use of the
    network medium
  • There are no collisions
  • The nodes must have the token in order to
    transmit
  • The network hardware ensures that there is only
    one token at any given time

43
Token-Based Networks
  • Advantages
  • No collisions, so no bandwidth is wasted by
    collisions and re-transmits
  • No need for re-collision avoidance schemes
  • Disadvantages
  • Token transmission uses bandwidth
  • More complicated hardware
  • Hardware must be built to use tokens, dynamically
    determine token sequence, etc.

44
Local Area Networks
  • Networks which span a small geographic area
  • They typically represent high bandwidth, short
    delays, few errors
  • They commonly support features such as
    broadcasting, multicasting
  • They are typically limited to hundreds of network
    nodes (maximum)

45
Typical Local Area Networks
  • A collection of computers in the same room
  • e.g. The basement of the computer centre
  • All computers within an office building
  • e.g. The computers in the offices of the
    professors and staff in Lambton tower

46
Local Area Network Topologies
  • Structures of LANs

47
Token Bus Networks
  • The token is passed in a specific sequence
  • Nodes must know the address if the next node in
    the sequence
  • The token sequence is not necessarily in the same
    order as the physical order of nodes on the
    communication medium
  • When a node has completed transmission, it
    forwards the token, addressed to the next node in
    the token sequence
  • The token sequence forms a logical ring

48
Common Token Bus Networks
  • IEEE 802.4 networks
  • Nodes are share a communication medium similar to
    that of Ethernet (IEEE 802.3)
  • Coaxial cable connection

49
Token Bus Operation
Transmit
A
C
B
D
Token sequence C,A,D,B
50
Token Bus Operation
Transmit Token
A
C
B
D
Token sequence C,A,D,B
51
Token Bus Operation
Receive Token
A
C
B
D
Token sequence C,A,D,B
52
Token Bus Operation
Transmit
A
C
B
D
Token sequence C,A,D,B
53
Token Bus Operation
Transmit Token
A
C
B
D
Token sequence C,A,D,B
54
Token Bus Operation
A
C
B
D
Receive Token
Token sequence C,A,D,B
55
Token Bus Operation
A
C
B
D
Transmit
Token sequence C,A,D,B
56
Token Ring Networks
  • The token is passed to each node, in the physical
    order on the network
  • The physical medium must be a closed loop to meet
    this network category
  • So the token can keep going around the network

57
Common Token Ring Networks
  • IEEE 802.5 networks
  • Nodes are share a coaxial communication medium
    similar to that of Ethernet (IEEE 802.3)
  • FDDI networks (fibre distributed data interface)
  • Nodes use 2 fibre optic rings as the
    communication medium
  • CDDI networks (copper dist. data interface)
  • Based on FDDI technology, but uses copper wiring
    similar to 802.4
  • However, CDDI uses 2 rings like FDDI

58
Token Ring Operation
A
Transmit
D
B
C
59
Token Ring Operation
A
Transmit Token
D
B
C
60
Token Ring Operation
A
Receive Token
D
B
C
61
Token Ring Operation
A
D
B
Transmit
C
62
Token Ring Operation
A
Transmit Token
D
B
C
63
Token Ring Operation
A
D
B
Receive Token
C
64
Token Ring Operation
A
D
B
C
Transmit
65
Bus and Ring Networks
  • Advantages
  • Less wiring is necessary
  • Disadvantages
  • Node failure can mean partial (or complete) LAN
    failure
  • This can mean locating network problems is also
    more difficult

66
Star Topology
  • Star networks send all messages through a central
    hub
  • Each node on the network is wired separately to
    the hub
  • Star networks are not a shared bus technology,
    but a private bus technology
  • However, nodes still share the hub

67
Common Star Networks
  • Twister pair Ethernet (logical star)
  • All nodes connect to a central hub (an Ethernet
    hub) via Cat5 cables
  • The hub forwards messages to all wires, and the
    destination node keeps the message
  • Other nodes ignore the message
  • An Ethernet switch (similar to an ATM switch)
    forwards only in the one correct direction (or
    not, if appropriate)

68
Star Network Operation
Transmit
A
B
Hub
C
D
69
Star Network Operation
A
B
Hub
C
D
Receive
70
Star Network Operation
Transmit
A
B
Hub
C
D
71
Star Network Operation
A
B
Hub
C
D
Receive
72
Twisted Pair Ethernet
  • Physically, all Ethernet types are bus networks
  • However, the actual layout of the cables in
    twisted pair Ethernet forms a star topology
  • Twisted pair is called a logical star topology,
    while still a physical bus topology

73
Twisted Pair Ethernet as a Bus
Short Shared Bus
B
C
A
D
Hub
F
G
E
H
Long Private Lines
74
Traditional Ethernet as a Bus
Long Shared Bus
B
C
A
D
F
G
E
H
Short Private Lines
75
Star Topology
  • Advantages
  • Simple installation and wiring
  • Node failures do not affect the rest of the
    system
  • Disadvantages
  • All traffic passes through same hub, so network
    bandwidth is limited by hub speed
  • This can be reduced with buffers inside hubs
    which store messages that come in when the hub is
    busy
  • Hub failure LAN failure
  • More wiring
  • Duplication of messages

76
LAN Service Models
  • In general, most LANs implement (in some sense)
    the OSI reference model
  • The IEEE committee on LAN technology (IEEE 802)
    chose to subdivide the Data Link Layer into 2
    sub-layers
  • MAC (Medium Access Control) Deals with issues
    specific to each type of LAN
  • Such as token passing, collision detection, error
    detection, etc.
  • LLC (Logical Link Control) Deals with issues
    common to all LAN types
  • Such as data transmission, etc.

77
Data Link Addressing
  • The data link layer is represents the network
  • e.g. Ethernet
  • Addressing, then, is specific to the network
    hardware
  • MAC addresses are typically used for this purpose
  • These addresses are not used in routing
  • They are only used on a single network
  • Thus, they are used for hop to hop delivery
  • End-to-end delivery is the domain of the Network
    layer

78
MAC Addresses
  • Officially the IEEE 802 committee standardized
    addresses to be 16bit, 48bit, and even 60bit
  • 48bit addresses (in use by most LANs covered by
    the 802 committee) allow for globally unique
    identifiers (GUIDs) to be assigned to each
    network card by the manufacturer
  • As a result, each NIC can be uniquely identified
    on any network
  • These are called MAC addresses, due to the Data
    Link sub-layer that deals with them
  • e.g. 8D-F0-A6-75-9C-13

79
Data Link Flow Control
  • Flow control is limiting the packet rate so that
    both the source or destination can keep up
  • At the data link layer, source and destination
    are on the same LAN
  • Thus, limiting the packet rate is relatively easy

80
Data Link Reliability
  • Reliability
  • Best effort The network takes no steps to
    ensure packets arrive
  • The majority of packets should be received
    without problems
  • Reliable The network uses acknowledgements to
    ensure packets arrive
  • When packets are lost (for whatever reason), they
    are handled appropriately
  • Error handling Corrupt packets should be
    re-sent
  • Reliability at the Data Link layer is usually
    unnecessary, since the Transport layer will
    typically be able to do it more efficiently

81
Error Control
  • Error control is achieved using one of the
    following methods
  • Checksum An n-bit sum is taken of the binary
    stream
  • In other words, a checksum counts the ones
  • What if one 0 became a 1 and a 1 became a 0??
  • Cyclical redundancy check
  • Should generate different CRC values, despite the
    same number of 0s and 1s

82
Ethernet
  • An Early Incarnation of LANs

83
What Started It All
Robert Metcalfe (from Xerox PARC)
84
Ethernet History
  • In 1973, Xerox PARC developed a packet-switched
    LAN, called Ethernet
  • In 1978, IEEE created a standard (802.3) based on
    the research of Xerox, Intel, and DEC
  • IEEE Institute of Electrical and Electronics
    Engineers
  • 802.3 Ethernet uses a coaxial cable to connect
    nodes (called 10Base5 or ThickNet)
  • Since then, several new forms of Ethernet have
    evolved

85
ThickNet (10Base5)
Outer Insulating Jacket
Inner Insulating Layer
Braided Metal Shield (Ground)
Transmission Wire
10Base5 5 gt 0.5
½ Inch Diameter
86
ThickNet (10Base5)
10Base5 10 gt 10 Mbps
  • Each network node uses a transceiver
  • A transceiver taps into the wire through holes
  • Maximum throughput is 10 million bits per second
    (10 Mbps)

Transceiver
87
ThinNet (10Base2)
  • Create as an inexpensive alternative to ThickNet
    (or 10Base2)
  • Called thin-wire Ethernet, because it uses a thin
    cable with less shielding
  • Less shielding means more interference, so cable
    placement is important
  • 10Base2 does not use transceivers, which are
    expensive, which further reduces cost

88
ThinNet (10Base2)
10Base2 2 gt 0.2
10Base2 10 gt 10 Mbps
Node A
Node D
Node B
Node C
  • The signal passes through each node
  • The network interface card (NIC) retransmits the
    signal, so transceivers are not required
  • Maximum throughput is 10 million bits per second
    (10 Mbps)

89
Twisted Pair Ethernet (10BaseT)
  • Uses 4 pairs of twisted wires inside an
    unshielded cable
  • The twisting of the wires reduces interference
  • The absence of shielding makes the cable flexible
    and inexpensive
  • The cable is capable of 10Mbps

90
Twisted Pair Ethernet
  • Connectors on twisted pair Ethernet (RJ45) look
    similar to telephone wire connectors (RJ11)
  • This kind of Ethernet uses unshielded twisted
    pair (UTP)
  • UTP cable has various categories
  • Cat3 Can only be used for 10BaseT
  • Cat5 Can be used for 10BaseT, 100BaseT
  • Cat5e, Cat6 Can be used for up to 1000BaseT

91
ThinNet Ethernet
011100110
011100110
92
Twisted Pair Ethernet
011100110
011100110
accept message
011100110
011100110
011100110
ignore
ignore
ignore
93
10 Mbps Ethernet Overview
  • 10Base2 and 10Base5 both used coaxial cable which
    joined each node in a line
  • 10BaseT uses UTP cabling, where each node is
    directly connected with the hub
  • The hub receives messages and forwards them to
    all nodes
  • The one that is connected to the recipient node

94
Fast Ethernet
  • Using the same Cat5 cabling used for 10BaseT, an
    Ethernet-based LAN that operates at 100 Mbps
    (100BaseT) is possible
  • Standard IEEE 802.3u
  • While using the same cable, network hubs and
    network interface cards (NICs) must be upgraded
    to transmit messages at 100 Mbps

95
Fast Ethernet
  • While very few computers can handle 100 Mbps
    throughput (bus speeds of computers are often
    slower than this), multiple computers can share
    this bandwidth
  • 10/100 Ethernet (or 10/100 switched Ethernet)
    allows you to use the same NICs and hubs for both
    10BaseT and 100BaseT
  • If a NIC and hub can both handle 100BaseT, that
    speed is used, otherwise 10BaseT is used
  • 10/100 Ethernet allows you to slowly upgrade your
    network with minimal downtime

96
Gigabit Ethernet
  • Gigabit Ethernet allows for 1000 Mbps throughput
  • Gigabit Ethernet (Gig-E) can use Cat5 cabling
    (1000BaseT) or shielded Cat5E cabling
    (1000BaseTX)
  • Standard IEEE 802.3ab
  • Gig-E pushes the limits of the speed capable with
    Cat5 cabling, due to interference with the
    electrical signal, Cat5E cabling results in
    better performance
  • Gigabit Ethernet is so fast, that it is sometimes
    used as a backbone for a Wide Area Network (WAN)
    instead of more expensive optical networks
  • e.g. One of the backbones of the network here at
    the U

97
Ethernet Future
  • Another form of Gigabit Ethernet which uses fibre
    optic cabling has been proposed (802.3z)
  • Using multimode (multiple channel 1000BaseSX),
    or single mode (1000BaseLH, 1000BaseZX)
  • Research groups are in the process of developing
    10 Gigabit Ethernet (802.3ae)
  • This research is managed by the 10 Gigabit
    Ethernet Alliance
  • http//www.10gea.org

98
LAN Service Models
  • LLC (Logical Link Control), for LANs, can be one
    of two types
  • Type 1 A straight datagram scheme
  • The packet is delivered using best-effort service
  • No acknowledgements are used to ensure packet
    arrival
  • Type 2 A reliable scheme
  • Packets are numbered
  • Packets are acknowledged as they are received

99
IEEE 802 Committees
  • Five 802 committees were developed to research
    various technologies associated with LANs
  • 802.1 Issues common to all LANs
  • e.g. addressing, management, bridges
  • 802.2 Issues related to the LLC sub-layer
  • e.g. reliability schemes, packet transmission
  • 802.3 Issues related to CSMA/CD category LANs
  • e.g. Ethernet
  • 802.4 Issues related to token bus category LANs
  • 802.5 Issues related to token ring category LANs

100
LAN Addresses
  • The 48 bit addresses (often called MAC addresses)
    are the ones used by Ethernet LANs
  • e.g. 02-60-8C-08-E1-0C
  • All Ethernet cards contain a globally unique MAC
    address

101
Ethernet Overview
  • Ethernet is not a reliable service
  • There are no acknowledgements for packet receipt
  • Ethernet uses best-effort delivery
  • Most Ethernet networks use broadcasting to
    achieve messaging
  • Each message is received by each node
  • Ethernet is one network in a category of networks
    known as shared bus networks
  • Each node shares a single communication medium

102
Ethernet Overview
  • Ethernet is a carrier-sensing network
  • Carrier-sensing networks use distributed access
    control methods
  • Each station determines whether it can access the
    communication medium
  • Each station senses whether or not the
    transmission medium (wire) is charged
  • If not, an attempt at transmission is made
  • If so, the node will wait and sense again

103
Ethernet Overview
  • Sometimes, more than one station will attempt to
    transmit at roughly the same time
  • This is called a collision
  • Due to the finite speed of electrons traversing a
    wire
  • 70 of the speed of light
  • Or due to the finite speed of photons moving
    through glass
  • The speed of light
  • The two (or more) messages collide or interfere
    with one another, creating scrambled data packets

104
Collision Detection in Ethernet
  • When scrambled messages are read by the
    transmitting stations, it is determined to be a
    collision
  • Both (or all) of the stations involved will
    detect the collision
  • This type of network is known as CSMA/CD
  • Carrier-sensing, multiple access with collision
    detection
  • Each station must retransmit their packets

105
Collision Avoidance in Ethernet
  • After a collision occurs, if both stations tried
    to transmit after the same period of time,
    another collision would occur
  • To combat this, Ethernet uses a binary
    exponential back-off policy
  • Each subsequent collision would cause the station
    to wait double the amount of time before
    reattempting transmission

106
Ethernet Packets (Frames)
  • Size 64 octets 1518 octets
  • An octet is another term for an 8-bit byte
  • The frame contains more than just data
  • The source and destination addresses
  • An identifier, signifying that the frame is in
    fact an Ethernet frame
  • A Cyclical Redundancy Check (CRC) to ensure data
    integrity upon arrival

107
Ethernet Frames
  • Sequence of 01010101 used to synchronize the
    receiving station
  • The MAC address of the destination node
  • The MAC address of the sender node
  • The identifier used to identify the frame as an
    Ethernet frame
  • The data to be sent to the destination
  • A cyclical redundancy check (CRC) used to
    determine if data has been corrupted

Preamble
8 octets 6 octets 6 octets 2
octets 46-1500 4 octets
Dest Address
Source Address
Frame Type
Data
CRC
108
Ethernet Distance Limitations
  • Coaxial Ethernet cables have a maximum length
  • Due to signal deterioration
  • This length could be extended using repeaters
  • Machines that read signals through a port and
    recreate them (at full strength) out another port
  • The use of more than 2 repeaters between any 2
    stations would interfere with times used in
    CSMA/CD schemes
  • As a result, a maximum of 2 repeaters can be
    placed between any 2 nodes

109
Ethernet Distance Limitations
  • Ethernet LAN sizes could also be increased by
    using Bridges to connect separate LANs into a
    single LAN
  • Bridges filter out erroneous frames, as well as
    line noise
  • Some bridges (adaptive bridges) are even
    intelligent enough to know when a frame must be
    forwarded or not
  • e.g. If the destination node is not on the other
    side of a Bridge, the frame need not be forwarded

110
FDDI
  • Fiber Distributed Data Interconnect

111
FDDI
  • Use optical fibre cabling as a shared
    communication medium
  • Optical fibre cables are made of glass
  • Because they are so thin, they are fairly
    flexible
  • Capable of 100 Mbps
  • Light is used to transmit data
  • Light is not susceptible to electrical
    interference
  • Optical cabling can span longer distances
  • Optical cabling does not need to be shielded near
    devices which generate electromagnetic
    interference
  • Light waves (photons) travel faster than electrons

112
FDDI
  • Is a token-ring category network
  • A token is passed from station to station
  • When a station receives the token, it may
    transmit data
  • If a station has no data, it allows the token to
    pass to the next station
  • FDDI uses 2 rings of cabling, moving in opposite
    directions
  • The second ring is used to allow twice the flow
    of data
  • The purpose of the second ring is to allow data
    to reach its destination, even when one station
    has failed (and cannot forward messages)

113
FDDI Ring Technology
114
FDDI With Node Failure
115
FDDI Token Passing
1
2
3
4
S12 D07
S12 D07
S12 D07
S12 D07
S12 D07
S12 D07
S12 D07
S12 D07
S12 D07
S12 D07
12
5
T
S12 D07
11
6
S12 D07
S12 D07
10
9
8
7
S12 D07
116
FDDI Token Passing
T
1
2
3
4
T
12
5
11
6
10
9
8
7
117
FDDI Frames
Preamble
Data Used to Synchronize Stations
octets 2
Start Delimiter
Indicates Start of Frame
1
Frame Control
Identifies the Type of Frame
1
Dest Address
Address of the Destination Node
2 or 6
Source Address
Address of the Source Node
2 or 6
Routing Info
Routing Information
0-30
Data
Frame Data
0-4500
FCS
Frame Check Sequence
4
End Delimiter
Indicates End of Frame
0.5
Frame Status
Status of Frame
1.5
118
Wireless Networks
  • Radio-Based LANs

119
Wireless LANs
  • Contrary to ones initial guess, wireless LANs
    are very similar to wired LANs
  • Wireless LANs are a shared media network, just
    like Ethernet
  • However, in a wireless LAN, the shared medium is
    not the air, but something called a base station
    or wireless access point

120
Wireless LANs (WLANs)
  • The wireless access point, which is similar to a
    hub, is the shared medium
  • Despite the fact that radio waves using the same
    frequency will cause mutual interference, the air
    is not generally considered a shared medium
  • Technically speaking, twisted pair Ethernet is
    similar to WLANs
  • The cables themselves are just point-to-point
    connectors and are not shared
  • The hub/switch, however, is shared

121
Wireless LANs (WLAN)
  • Wireless Access Point (WAP) A base station that
    coordinates transmission between one or more
    wireless hosts
  • Analogous to a cell tower in a mobile phone
    network
  • Wireless hosts must be a certain distance away
    from a WAP to participate on a WLAN
  • The communicable area of all of the WAPs in a
    WLAN, define the coverage area for the WLAN
  • Some WLANs do without a WAP, but pass messages
    directly to one another
  • These are typically small (2-3 hosts) networks,
    and are called ad hoc networks

122
802.11 Operation
  • 802.11 networks (such as 802.11g) use CSMA/CA
    multiple access scheme
  • Hosts try to detect carrier before sending (CS)
  • This is not adequate, since there could be hidden
    hosts
  • These are hosts out of range of this host, but in
    range of the same base station

123
802.11 Operation
  • To avoid collisions with hidden hosts
  • The host will send a request to send (RTS)
    frame before transmitting
  • The base station will respond with a clear to
    send (CTS) frame if the channel is clear
  • Once a base station sends a CTS, it will reject
    any further RTS requests until the data is
    received by the host who sent the first RTS
  • This is called collision avoidance (CA)
  • Frames are acknowledged at the data link layer in
    802.11 networks

124
802.11 Frame Format
  • Flags
  • MAC Address of sending host
  • MAC Address of receiving host
  • MAC Address of sender base station
  • Fragment number, sequence number
  • MAC Address of receiver base station
  • Frame data
  • CRC for frame header and data

Frame Control (2 octets)
Source Address (6)
Destination Address (6)
Receiving Station Address (6)
Sequence Control (2)
Transmitting Station Address (6)
Data (0-2312)
Frame Check Sequence (2)
125
802.11 Frame Header Frame Control
Protocol Version (2 bits)
  • Flags
  • Management, control or data frame
  • Type of management or control frame
  • Sent to an access point?
  • Sent by an access point?

Type (2)
Subtype (4)
To AP (1)
From AP (1)
More Fragments (1)
Are there more fragments from this frame? Is this
a retransmission of a previous frame? Power state
of sender after transmission Is there more data
to come? Has WEP encryption been applied to
frame? Are the packets strictly ordered?
Retry (1)
Power Management (1)
More Data (1)
WEP (1)
Order (1)
126
Wireless Access Points
WAP2
WAP1
WAP3
127
Ad Hoc Networks
  • In ad hoc networks, stations directly transmit to
    one another
  • Hosts are responsible for routing, addressing,
    name translation, security, etc.
  • Two ad hoc networks using the same frequency,
    within range of one another will cause conflicts
  • Thus, different frequencies should be used

128
Handoffs in WAPs
  • For WLANs with WAPs, roaming hosts must be
    considered
  • If a host moves into the range of another WAP,
    then out of range of their current WAP, a handoff
    takes place
  • A handoff is when one WAP gives the
    responsibility for a particular host to one of
    its neighbouring WAPs
  • The two WAPs must communicate for this to happen,
    and thus neighbouring WAPs must be within each
    others transmission range

129
Wireless LAN Standards
  • Some of the main standardized WLANs
  • 802.11a, 802.11g 54Mbps, comparable with
    100BaseT Ethernet, under 100M range
  • 802.11b 11Mbps, comparable to 10BaseT Ethernet,
    under 100M range
  • These technologies are intended for LANs within
    the same small to medium-sized building
  • BlueTooth/802.15 721 kbps, under 10M range
  • This technology is intended for communicate
    within one room or vehicle
Write a Comment
User Comments (0)
About PowerShow.com