Transcript Advanced Review (Part II)

Local Area Networks
Content
Chapter 14: Advanced Review (Part I)
Switched Ethernet
Ethernet Evolution
Shared vs. Switched LANs
Transparent Learning
Spanning Tree Protocol
Tutorial Questions
Additional Notes
Ethernet Evolution




Developed in the mid-1980s as a shared bus LAN
– Operated over coaxial cable
– Used CSMA/CD channel access algorithm
Repeater and hubs (Layer 1 relays) extend distance
– Number restricted to four between any
two nodes on 10Mbps Ethernet
Bridges (Layer 2 relays) overcome restrictions
on number of repeaters
– Spanning Tree Protocol (IEEE 802.1D)
addresses bridge resilience issues
Twisted pair cabling introduced in late 1980s
– Reduced network diameter
– More resilient than physical bus
– Ethernet hub replaced repeater
Hub
CSMA/CD = carrier sense multiple access with collision detection
Repeater
or bridge
Bridge
Hub
Ethernet Evolution
(continued)


FOIRL = fiber optic inter-repeater link
UTP = unshielded twisted pair
Fast Ethernet standardized by mid-1990s
– Supported on legacy UTP-3 and upcoming UTP-5 cabling
– Reduced network diameter compared to 10Base-T
Auto-negotiating 10/100 interfaces self-configure speed and duplex mode
– Flow control prevents overrun on 10 Mbps interfaces
Data
rate
Cable types and distance limitations (meters)
Coaxial cable
(10Base5 and
10Base2)
UTP-3 4-pair or
UTP-5
(10Base-T and
100Base-TX)
MMF
(10Base-F and
100Base-FX)
10
Mbps
500 (no
repeaters)
2500 (max. 4
repeaters)
100 (no repeaters)
500 (max. 4
repeaters
2000–3000 using 1
km FOIRL
100
Mbps
—
205
2000
Gigabit Ethernet (GbE)


STP = shielded twisted pair
Gigabit operation standardized in 1998
– After Fibre Channel became established
 Used Fibre Channel physical layer chips for 1-Gbps duplex operation
– Needs fiber or enhanced quality copper (UTP-5e or better)
Shared (CSMA/CD) GbE added in 1999
– Restricted to single hub
– Extended Ethernet collision window to 4096 bit times (4.096µsec)
 Introduced compatibility issues with 10Base-T and 100Base-T
–
Has proved unpopular compared to switched GbE
Data rate
Distance limitations (meters)
STP 4pair
1000
Mbps
25
UTP 4pair
MMF
SMF
100
50/125:
up to
550
62.5/125
up to
440
10,000
Link Aggregation

Initially, a proprietary switch hardware feature
–

Now standardized as IEEE 802.3ad
–

Further modifies LAN Layer 2
Example
–
–

Available on Fast Ethernet and Gigabit Ethernet
interfaces
New!
LLC
Link Aggregation
MAC
control
MAC
…
PHY
Aggregating four 100-Mbps inter-switch links
gives aggregated bandwidth of 400 Mbps
Is cheaper than upgrading switches to GbE
Link Aggregation Control Protocol (LACP) negotiates
load sharing
–
Fat pipes are treated as a single link by Spanning Tree
ProtocolLLC = logical link control
MAC = media access control
PHY = physical
Fat 400 Mbit/s pipe
MAC
control
MAC
PHY
10 Gigabit Ethernet (10GbE)






Standardized in 2003 as IEEE 802.3ae
Initially aimed at MAN/WAN links and storage area networks
– Not designed for use in LAN
Switched full duplex only
– CSMA/CD neither supported nor required
Current standard only supports optical fiber
– Copper versions under investigation
 Poses new engineering challenges!
Framing is compatible with earlier versions of Ethernet
Novel additions include
– WAN interface definition for connecting to SDH/SONET MANs and WANs
– New, high-specification multimode fiber type and PHY-PMD interface
MAN = metropolitan area network
PMD = physical medium dependent
10GbE Architecture
Full duplex MAC
XGMII or XAUI
WWDM PHY
(8B/10B coding)
Serial LAN PHY
(64/66B coding)
Serial WAN PHY
(64/66B coding + WIS)
WWDM PMD
(1310nm)
Serial Serial
Serial
PMD
PMD
PMD
850nm 1310nm 1500nm
Serial Serial
Serial
PMD
PMD
PMD
850nm 1310nm 1500nm
10GBASE-CX4
10GBASE
-LR
-SR
WIS = WAN interface sub-layer
WWDM = wideband wave division multiplexing
-ER
10GBASE
-LW
-SW
-EW
XAUI = 10Gbps attachment user interface
XGMII = 10Gbps medium independent interface
10GbE Operating Distances



10GbE specification supports four fiber types
– Two MMF types and two SMF types
Also allows three wavelengths: 850, 1310, and 1550 nm
Leads to a number of operating distances!
– Table shows selection of distances (in meters)
 Shorter distance is with 62.5/125-µ cable, longer one with 50/125-µ
cable
Wavelength 
Fiber type

850 nm
1310 nm
1550 nm
10GBASE-S
26–300
―
―
10GBASE-L
―
10,000
10GBASE-E
―
―
30,000–
40,000
10GBASE-LX4
300
300
10,000
Switched Ethernet
Ethernet Evolution
Shared vs. Switched LANs
Transparent Learning
Spanning Tree Protocol
Tutorial Questions
Additional Notes
Hubs, Bridges and Switches

Hubs extend the collision domain
–
They are Layer 1 devices


Bridges and switches are both
Layer 2 LAN devices
–

Operate on MAC frames
Early bridges had limited connectivity
–

Operate on bits
Often operated on one frame at a time
Switches have considerable connectivity potential
–
Can operate on several frames in parallel
Full Duplex Operation



Traditional CSMA/CD operates in shared (half duplex) mode
– Host can either write to network or read from network
 But not do both at same time
Traditional CSMA/CD hubs have CSMA/CD LAN ‘inside the box’
– Therefore are inherently half duplex
Extension to standard allows simultaneous reading and writing
– Full duplex operation
– Conditional on being supported by network device and host NIC
 And there being only one NIC attached to device port
Conventional hubs:
inherently half duplex
NIC = network interface card
Systems on dedicated switch ports:
could be half- or full duplex
Auto-Negotiation Protocol






Modern CSMA/CD NICs available in two main forms
– 100/100 or 10/100/1000
Notation means they are capable of selecting
– Speed at which they operate (10, 100 or 1000 Mbit/s)
– And the mode at which they operate
 Half- or full-duplex
Selection process carried out by an auto-negotiation protocol
– Runs between NIC and switch
– Also works with some hubs
Protocol attempts to negotiate highest throughput first
– 1000Mbit/s or 100Mbit/s full-duplex operation
Then works down through list to lowest throughput
– 10Mbit/s half-duplex
Full duplex operation is collision-free
Bridges and Switches I

Bridges and LAN switches are MAC Layer relays
–
–
–
Used to interconnect LANs of same type
Use LAN MAC addressing
Bridge/LAN switch
Operate on LAN frames
802.3,5,11, etc
Physical: to match
Data Link Protocol
Layer 2
relay
Bridges and Switches II


Each port connects to a different collision domain
– Supports parallel activity on each attached LAN
segment
Transparent to routers and host operating systems
layers 5/6/7:
Application
layers 5/6/7:
Application
TCP, UDP
TCP, UDP
IP
IP
802.3,5,11,..
802.3,5,11,..
802.3,5,11,..
802.3,5,11,..
802.3,5,11,..
Physical
Physical
Physical
Physical
Physical
Shared LANs





Shared LANs developed for
– File and print sharing (often on local servers)
– Networked applications (may be on remote servers)
– Bandwidth sharing, where dedicated bandwidth not required
 Or dedicated bandwidth too expensive to justify for one system
LANs have used structured cabling systems since early 1990s
– High specification cable interconnecting ‘wiring closets’ and network
devices
– Facilitate high transmission rates
Many of today’s wired LANs are CSMA/CD
– Operating at 10/100 or 10/100/1000Mbit/s
– Fully switched LANs increasingly popular
Wireless LANs (11 and 54 Mbit/s) also becoming common
– Operate as shared media LANs
 Discussed later in course
Switched LANs offer significant throughput increase over shared LANs
Switched vs. Shared Bandwidth

Example: Twelve users and four
servers share 100Mbit/s LAN
–
–

All in same collision domain
Access time to shared channel
increases as usage increases
Solution to increasing congestion:
replace shared LAN with 10/100Mbit/s switch
–
Users divided into smaller
collision domains

–
Each receives larger portion
of bandwidth
Switch throughput at least
port speed  ½ number of ports

Eight-port switch supports
up to 400Mbit/s throughput
Switched LANs

Switches commonly used for LAN-LAN interconnection
–
Usually interconnect same technologies

–

10, 100, 1000 and 10000Mbit/s
Rapidly decreasing support for older technologies
–

Falling switch prices mean end of hub market
Switches available for all versions of CSMA/CD
–

For example, CSMA/CD to CSMA/CD
Token Ring (16 & 100Mbit/s), FDDI and ATM (25, 155 &
622Mbit/s)
CSMA/CD switches have become widespread due to
–
Their versatility

–
Support of different bit rates and media types
Their lower per-port cost than alternative technologies
Ethernet Switches

Began to appear in mid-1990s as combined
hub/bridges
–

Now commodity items
–

Multiple 10/100 interfaces at very low per-port cost
Unmanaged switches have become cost-efficient hub
replacements
–

Lower latency and lower per-port cost than bridges
Many manufacturers no longer make hubs
Managed switches have CPU, memory and multitasking
operating system
–
Support many additional features




VLANs
QoS
Multicast
And, of course, network management
QoS = quality of service
VLAN = virtual LAN
Ethernet Switches
(continued)
Late 1980s
Bridge
Hub
Hub
Mid-1990s
Late 1990s
Switch
Switch
Hub
Hub
Multilayer Switches

Multilayer switches have both switching and
routing modules
–
–
Operate at Layer 2 and Layer 3
Often very high-speed and expensive devices

–
Typically equipped with hardware acceleration
Used in backbone (or ‘distribution’) networks
Multilayer switch
Modern LAN Structure

Workgroups
connected to small
switches
–
–

Workgroup servers
get dedicated ports
10 and 100Mbit/s
connections
Wiring
closet
–
The backbone or
distribution network
100 and 1000Mbit/s
connections used
Workgroup
servers
Fiber
links
Hub
Workgroup switches
interconnected by
multilayer switches
–
10/100
switch
Hub
Hub
Multilayer
switch
Site backbone
Gigabit Ethernet/ATM
Switched Ethernet
Ethernet Evolution
Shared vs. Switched LANs
Transparent Learning
Spanning Tree Protocol
Tutorial Questions
Additional Notes
Standards and Terminology




Bridge(switch) interfaces conform to specific IEEE LAN standard
– For example, IEEE 802.3 Ethernet family
And operate according to the IEEE bridging standards
– 802.1D: CDMA/CD transparent bridging and Spanning Tree
– 802.1Q: Virtual LANs (VLANs)
– 802.1p: Traffic classes and prioritization
The term switch is not standardized
– Used in different ways to suit manufacturers’ marketing policies
 For example, ‘multilayer switch’
The term bridge defines a generic MAC-layer relay
– Literature uses this term
Relay
logic
– We shall also use it in when discussing
MAC1
MAC2
these devices in the context of a standard
Phys1
Phys2
Transparent Learning Bridges



Each interface operates in promiscuous mode
B
G
– Receives and processes all frames from LAN
attached to that interface
1
Devices build a MAC address table
C
T
A
– Inspect each frame header
2
– Record frame’s source MAC
address (MACSA) and the port
number on which the frame
entered
N
K
B
Process is called
transparent learning
MAC address table
– Determines how data
Port Port
frames processed
IP
=
A
MACSA = A
Upper layer SA
1
2
IP
=
K
MACDA = K
subsequently
DA
info
MAC
MAC
A
N
MAC
MAC
G
K
The Three F’s

Bridge extracts MACDA from incoming frame
headers and looks it up in MAC address table
–

B
To ensure frame reaches correct LAN segment
G
Filters frames between systems on same port
–
–

Device then makes forwarding decision
Forwards frame to systems on different ports
–

B  G: filtered
Since systems on same port, normal MAC-level
addressing ensures that frame reaches
destination
1
A
C
T
No further action required by bridge
2
Floods broadcast and multicast frames
–
Bridge also required to flood frames with
unknown MAC destination addresses
–
Flooding means sending a copy of the
incoming frame to all other ports
H
K
A  K: forwarded
N
T  unknown:
flooded
Static Filtering



Older bridges could be programmed with static filters
Examples
–
Filter frames to system G arriving on port 1
–
Filter AppleTalk frames arriving at port 2
This function eventually taken over by routers
B
G
All G: filtered
1
A
C
T
All AppleTalk: filtered
H
K
2
N
MAC Address Tables

Identify the port on which known systems
can be reached
–
Could be via other switches
B
C
G
H
Port
1
Port
2
Port
3
A
G
L
B
H
M
C
N
R
A
SW1
S
2
1
T
3
3
2
1
SW2
T
L
S
Port
2
Port
3
L
R
A
M
S
B
N
T
C
G
H
N
M
Port
1
R
Switched Ethernet
Ethernet Evolution
Shared vs. Switched LANs
Transparent Learning
Spanning Tree Protocol
Tutorial Questions
Additional Notes
Single Points of Failure

A single bridge interconnecting two LANs
constitutes a single point of failure
–
More than one user affected if device fails
LAN 2
B
G
1
C
A
LAN 5
T
2
LAN 1
B
K
N
LAN 3
LAN 4
What About Using Extra
Bridges for Resilience?





Bridges learn MAC addresses as usual
– For example, a frame from H to P
Each bridge queues a copy of the frame for forwarding to
the other LAN segment
Other port on each bridge receives copy of frame
– Notes new port for H
– Queues frame for forwarding to original LAN
– And looping starts to occur
 … but when does it stop?
Suppose you were to use two extra bridges
instead of just one?
– Or there is more than one loop?
Bridges use the Spanning Tree Protocol
to resolve the problems of looping frames
Q
B1
H
1
2
G
P
B2
1
F
N
2
The Spanning Tree Protocol
(IEEE 802.1D)




Supported by all MAC-Layer bridges (switches)
– Runs automatically (unless disabled)
Bridges send each other topology messages
(‘BPDUs’) to build the spanning tree
– A loop-free topology
When the protocol has run, certain bridge ports
will forward MAC data frames
– Other ports will be blocked *
– Leaves a single path between any two LANs
 Inter-switch links treated as LANs in this
context
Once configured, protocol periodically re-affirms
topology
– Reconfigures spanning tree if topology fails
* All ports continue to receive BPDUs, even if blocked for data frames
B1
H
1
Q
2
G
P
B2
N
1
2
F
BPDU = bridge protocol data unit
The Spanning Tree Protocol
(IEEE 802.1D)
Motivation for Spanning Tree




A collection of LANs that are interconnected by a set of bridges,
The interconnection may have redundant to increase reliability,
Redundancy introduce loops which defeats the objective because multiple reception of same
packets (diff. routes) and routing back to source are possible,
A spanning tree create a single virtual route, although multiple physical routes are there.
Setting up a Spanning Tree






The bridge (B) with lowest ID is selected as Root Bridge (RB),
The root port (RP) of B, other than RB, is a port of B such that: the link(RP,RB) is least among
all ports of B, where least cost means minimum number of hops, least delay, or maximum
bandwidth,
There should be one designated B (DB) for each LAN-X. DB is defined as a B directly connected
to LAN-X such that Path(LAN,RB) is least is route is done through DB.
There should be a designated port (DP/DB) connecting DB to LAN-X, if ties then lowest ID.
Make Root Ports (RPs) and Designated Ports (DPs/DB) as forwarding ports and all the other as
blocking ports.
The results: packets flowing through RPs and DPs/DB follow a spanning tree route without
loops.
Spanning Tree Overview



Spanning Tree operation based on two device
parameters
– Bridge identifier, or BID
– Notional cost of leaving bridge on given
port
Bridges flood BPDUs to determine bridge with
lowest BID
– This bridge becomes Root Bridge
Other bridges then
1. Identify their root ports
 Those with lowest cost path to Root
Bridge
2. Identify any designated ports
 Those responsible for forwarding
frames away from the Root
3. Block their non-designated ports
 Those with higher-cost path to Root
H
G
bridge 1
priority 10
1
2
10
15
Q
P
bridge 2
priority 1
F
1
2
10
5
Port costs
N
Bridge Identifiers

Two fields concatenated into 64-bit number
–
16-bit priority and 48-bit MAC address


Priority is at most significant end of number
Standard recommends default of 32768
Reducing priority increases bridge’s Root eligibility
–

Sending port’s
MAC address
Manufacturers normally select default value for priority
–

Priority
For example, reducing the priority to 32700
MAC address acts as tie-breaker if all priorities equal
MAC = 0060.6475.6bc0
MAC = 0060.6475.6b00
MAC = 0060.6475.6d05
Port cost


Every bridge/switch port has outgoing port
cost
– Older bridges used 109/bandwidth by
default
– Newer versions use non-linear scale of
IEEE 802.1D
– Port costs can usually be set manually
Root path cost is the sum of (outgoing)
path costs to Root from current bridge
– Cost is 0 for all Root ports
Root bridge
100
Root path
cost = 138
19
19
100 Mbps
Outgoing
cost = 19
100 Mbps
10 Mbps
Outgoing
cost = 100
10 Mbps
Bandwidth
Cost
10 Mbps
100
16 Mbps
62
45 Mbps
39
100 Mbps
19
155 Mbps
14
622 Mbps
6
1 Gbps
4
10 Gbps
2
LAN Switching
Hubs and Media
Bridges and Switches
Transparent Learning
Spanning Tree Protocol
Spanning Tree Examples
Note on VLANs
Two-Switch Example
Stage I: Root Discovery



Both bridges flood BPDUs with
Root Bridge = self for a preset period
At end of this stage, B2 found to have
lowest BID
– Priorities same (32768)
– B2’s MAC address lower than B1’s
B2 becomes Root Bridge
B1
Priority 32768
0060.6475.6bc0
H
1
2
G
P
B2
Priority 32768
0060.6475.6b00
B1’s BID 32768
0060.6475.6bc0
1
2
F
B2’s BID 32768
Q
0060.6475.6b00
LAN 1
(10Mbps)
LAN 2
(10Mbps)
N
Stage II: Find lowest cost root path



Non-root bridge(s) determine
lowest-cost path to root
Equal-lowest-cost paths decided by
series of tie-breakers
1. Upstream bridge with lowest BID
 Where more than one bridge
available
2. Lowest port ID on bridge
 Where bridge has lowest
cost path from more than one port
 Defaults to lowest port number
B1 has two lowest cost paths to root
– Port 1 is lower port number than port 2
 B1 port 1 becomes root port
– Alternate path via Port 2 will be blocked
 Becomes a non-designated port
B1
Priority 32768
0060.6475.6bc0
H
1
Q
2
G
Root path
Cost = 100
Root path
Cost = 100
B2
Priority 32768
0060.6475.6b00
1
2
F
LAN 1
(10Mbps)
LAN 2
(10Mbps)
P
N
Stage III: Finalize Spanning Tree



B1
Root Bridge
Priority 32768
– Places all active ports in forwarding
0060.6475.6bc0
H
state
1
2
 All active ports on root bridge
X
are designated ports
RP(F)
NDP(B)
Non-root bridges modify port states
G
– Root port placed into forwarding state
 B1, port 1
– Designated port(s) placed into
forwarding state
B2
Priority 32768
 In this example, B1 has no
0060.6475.6b00
designated ports
– Non-designated ports moved to blocking
1
2
state
DP(F)
DP(F)
F
 B1, port 2
Root bridge periodically sends out Root
LAN 1
LAN 2
BPDU to reaffirm topology
(10Mbps)
(10Mbps)
Q
P
N
Further Example: Before
LAN 2
C = 10
C=5
Bridge 4
C=5
Bridge 3
C = 10
C = 10
Bridge 1
C = 10
LAN 5
C=5
Bridge 5
C=5
C=5
C = 10
Bridge 2
LAN 3
C=5
LAN 1
LAN 4
Further Example: Before
LAN 2
C = 10
C=5
Bridge 4
C=5
Bridge 3
C = 10
C = 10
Bridge 1
C = 10
LAN 5
C=5
Bridge 5
C=5
C=5
C = 10
Bridge 2
LAN 3
C=5
LAN 1
LAN 4
Stage I: Identify Root Bridge
LAN 2
C = 10
Bridge 3
C = 10
C = 10
Bridge 1
C = 10
LAN 5
C=5
Bridge 5
C=5
LAN 1
C=5
C = 10
Bridge 2
LAN 3
C=5
Bridge with
lowest BID
becomes
Root
C=5
Bridge 4
C=5
LAN 4
Stage II: Identify Forwarding Ports
LAN 2
RPC = 10
C = 10
RPC = 15
Bridge 3
C = 10
C = 10
RPC = 0
Bridge 1
C = 10 RPC = 0
LAN 5
C = 5 RPC = 10 or 15
Bridge 5
RPC = 5
C=5
C = 10
Bridge 2
LAN 3
C=5
LAN 1
C=5
Notes
1.
RPC = root path cost
2.
Bridge 2 has only one
root path
3.
Bridges 4 & 5 tie on
lowest path to Root for
LAN 5
– Lowest BID
is tie-breaker (4)
4.
All Root Bridge ports
are designated ports
with RPC of zero
C=5
RPC = 5
Bridge 4
RPC = 10
C=5
(No path to
root)
RPC = 10
LAN 4
Stage III: Finalize Spanning Tree
LAN 2
The Spanning Tree
L1
B2
L4
B1
C = 10
L2
B5 B3
L3
Bridge 3
C = 10
B4
L5
C=5
Bridge 4
C=5
C = 10
Bridge 1
C = 10
X
LAN 5
X
C=5
Bridge 5
C=5
LAN 1
C=5
= Forwarding
X
C = 10
Bridge 2
= Blocking
LAN 3
C=5
Key
LAN 4
Summary

Variety of media are used on CSMA/CD LANs
– But the most common media today are twisted pair and optical fibre

LAN switches have replaced bridges
– Behave exactly like bridges
– But have higher connectivity and throughput

Bridges and switches divide LANs into separate collision domains
– But interconnected LANs are still a single broadcast domain

Modern LANs must have some degree of fault tolerance
– Provided by installing additional switches and links

CSMA/CD bridges use the Spanning Tree Protocol to create a
loop-free topology that spans whole Layer 2 domain
Virtual LANs provide additional traffic control in switched LANs

Switched Ethernet
Ethernet Evolution
Shared vs. Switched LANs
Transparent Learning
Spanning Tree Protocol
Tutorial Questions
Additional Notes
Tutorial Questions
1.
At which OSI layers do bridges and switches operate?
2.
What is meant by a transparent learning bridge?
3.
How, and why, are frames to unknown destination
addresses treated like broadcasts and multicasts?
4.
What is the reason for using Spanning Tree and why is it
required?
5.
Full duplex operation is likely to be most beneficial for
what types of host?
Tutorial Questions
(continued)
6.
In the diagram below, there are three hosts systems, A, B & C and one server,
D all connected to 10Mbit/s LAN segments
6. Briefly describe how a frame from A reaches D, assuming that all systems have just been
switched on; include a description of how the ARP from A is processed by the bridges.
7. Show the entries of the port tables in bridges B1 and B2, once the location of all three systems
have been determined.
A
LAN 2
(10Mbit/s)
B1
LAN 1
(10Mbit/s)
B
C
B2
LAN 3
(10Mbit/s) D
Tutorial Questions
(continued)
7.
Here is the LAN diagram again, but an additional bridge, B3, has been
connected as shown, requiring the use of Spanning Tree. (The diagram shows
the MAC address of each bridge, all of which have equal priority.)
Re-draw the diagram showing the Root Bridge and labelling all bridge ports
as root ports, designated ports or non-designated ports and showing which
ports are forwarding and which are blocking.
Show one way in which you could connect the above LANs and server, with
another LAN segment and three further servers into a single eight-port
switch.
A
B1
0000.0c07.ac01
LAN 1
(10Mbit/s)
B3
0006.28c3.03c0
B
LAN 2
(10Mbit/s)
B2
0004.2875.c860
LAN 3
(10Mbit/s)
C
D
Switched Ethernet
Ethernet Evolution
Shared vs. Switched LANs
Transparent Learning
Spanning Tree Protocol
Tutorial Questions
Additional Notes
Repeaters and Hubs



Operate at the Physical Layer
– Physical Layer relays
– Unit of transfer is the bit
Extend domain of MAC protocol
– The collision domain
– Repeat incoming bits to other
ports
 MAC frames seen by all
systems
 Systems contend for
extended communication
channel
Support a variety of media types
–

Allows old style shared coaxial
segments to be connected to
modern twisted pair segments
Most hubs just multi-port
repeaters
Phys
relay
logic
1
Phys
2
hub
Coaxial LAN
segment
relay
logic
Ph1 Ph1 Ph1 Ph1 Ph1 Ph1 Ph1
10BaseT Hubs


Have separate 10BaseT ports for each
system
–
Enhances LAN resilience
–
Works in conjunction with structured
cabling systems
–
Can be cascaded to interconnect multiple
LAN segments
Maximum number of hubs/repeaters allowed
between any two systems varies with media
type and bit rate
–
10BaseT

–
No more than 4 (same as for coax
cable)
100BaseT

No more than 2 with twisted pair
cable
Coaxial LAN
segment
CSMA/CD Media Types & Limitations

LAN segment lengths limited by two factors
–
The operation of the CSMA/CD protocol
–
The media type


CSMA/CD collision window sets maximum amount of time for
detecting a collision
–
Specified by the 10Base5 standard as 51.2µs at 10 Mbit/s

–

Strictly, the bandwidth of the media
Equal to 512 bit times
At 100Mbit/s the collision window becomes 5.12µs (still 512 bit
times)
Different media have different transmission qualities
–
Structured cabling systems specify maximum distance from
wiring closet to desktop system of 100 metres

Standards committees meet this criterion for all desktop
LAN speeds
CSMA/CD = carrier sense multiple access with collision detection
CSMA/CD Media Types & Limitations
(continued)
Max.
Number
of
stations
per cable
Media
Type
Data rate
(Mbit/s)
Max.
cable
length
(metres)
Twisted
Pair
10, 100,
1000
100
Two
Thin
Coax.
10
185
30
Thick
Coax
10
500
100
Optical
Fibre
10, 100,
1000,
10000
Depends
on fibre
type and
data rate
Two
Evolving Technologies for
Ethernet LAN Interconnection
1980 –
1984
Shared Ethernet deployed
International LAN standards developed
1985 –
1989
Bridges used for LAN interconnection to limit size of
collision domains, with Spanning Tree facilitating bridge
redundancy; routers used for LAN-WAN interconnection
1990 1994
High-speed, low-cost routers become alternatives to
bridges
‘Backbone routers’ developed for site interconnections
1995 1999
VLAN-capable switches replace bridges and LAN routers
100Mbit/s Ethernet becomes common, GbE developed
2000 -
Switched access and VLAN deployment become common
10GbE developed, Ethernet switches become QoSenabled
The Rise and Fall
of the LAN Router




In early 1990s, small routers introduced to limit size
of broadcast domains
– Became cheap, and fast, enough to use in LANs
But routers operate at Network Layer
– Require configuration (are not plug-and-play)
– Have higher per-port cost than equivalent bridge
LAN switches began to replace bridges in mid-1990s
– Still operate at Layer 2
– Have much lower per-port cost than routers
– Can be operated in plug-and-play mode or configured
 For example with management and VLAN information
Routers still required for inter-site and inter-VLAN
communication
– Particularly suitable for interconnecting different
technologies
 For example, CSMA/CD & Frame Relay, CSMA/CD &
Token Ring