Transcript Repeaters and Bridges

Extending LANs
Based on Chapter 11 in Computer
Networks and Internets
CSIT 220 (Blum)
1
Size Limitations
• Recall that the combination of protocol and
wiring scheme places size restrictions on a
LAN
– Ethernet using 10Base5 (thick) allows segments
up to 500 m
– FDDI (Fiber Distributed Data Interface) allows
a single ring to be 100 kilometers (62 miles)
CSIT 220 (Blum)
2
Reason One
• Attenuation and interference:
– The farther a signal travels, the more it
attenuates (weakens). If the amplitude is
reduced, the signal-to-noise ratio may decrease.
And the information may be lost in the noise.
– The farther a signal travels, the more likely it is
to experience interference (noise) — other
unwanted waves adding to it.
– Increasing the amplitude would help but that
requires additional power and thus increasing
cost.
CSIT 220 (Blum)
3
Reason Two
• Sharing Speed:
– The longer a token ring (and the more
computers on it), the longer it takes before it is
a given computer’s turn to transmit.
– The longer a bus (and the more computers on
it), the more likely that two computers will
transmit simultaneously causing a collision
(CSMA/CD).
CSIT 220 (Blum)
4
Fiber-Optic Extensions
• The attenuation/interference limitation can be
overcome by changing the medium: wire  fiber
optic cable.
• Recall fiber optic cable has a much higher
bandwidth and is much less susceptible to
interference.
• Building a LAN entirely of fiber is expensive.
– One of the reasons FDDI did not become more
widespread is because of its cost.
• But fiber optic cable can be used to extend certain
connections in a LAN.
CSIT 220 (Blum)
5
Fiber Optic Extensions (Cont.)
• The connection between the computer and the bus
could be partially spanned by fiber optic cable.
• Electrical signal goes to a fiber modem.
• The fiber modem converts the signal into an
optical form for transmission over the optical
fiber.
• A second fiber modem receives the optical signal
and converts it back to an electronic signal.
• The electronic signal is placed on the bus in the
usual way.
CSIT 220 (Blum)
6
Fiber Extension (Fig. 11.1)
Note as shown above one is adding a single computer
that is a large distance from the rest of the network. But
this idea can be extended.
CSIT 220 (Blum)
7
Repeater
• Another solution to the attenuation/interference
problem is to use a repeater.
• A repeater receives a signal, amplifies it, and then
retransmits (forwards) it.
• Repeaters are used between segments of a local area
network (LAN).
– Segments have a fixed maximum length.
• Repeaters are also used in WANs, both wired and
wireless.
CSIT 220 (Blum)
8
Repeater (Cont.)
• In its simplest version an analog repeater
simply amplifies the signal it receives.
• Therefore, if any noise has corrupted the
signal, that noise is also amplified. This is
the best a repeater can do with an analog
signal.
• It is possible for a repeater to clean up a
digital signal provided it is not too noisy.
CSIT 220 (Blum)
9
It’s possible to clean up a digital signal
CSIT 220 (Blum)
A noise ridden digital signal.
10
It’s possible to clean up a digital signal
CSIT 220 (Blum)
A cleaned-up digital signal.
11
Repeater (Cont.)
• While noise can in some sense be removed
from a digital signal by a repeater, digital
signals tend to require more frequent
repeating.
• Whereas analog signal amplifiers are spaced
at 18,000 meter intervals, digital signal
repeaters are typically placed at 2,000 to
6,000 meter intervals.
CSIT 220 (Blum)
12
Repeater (Fig. 11.2)
CSIT 220 (Blum)
13
Repeaters
• Repeaters allow the LAN size to be extended.
• But the repeater solves only the attenuationinterference problem and not the sharing speed
problem.
• For this reason, some protocols place restrictions
on the number or arrangement of repeaters.
• Ethernet standards allow only a maximum of 4
repeaters between any two nodes.
CSIT 220 (Blum)
14
Multiple Repeaters (Fig. 11.3)
CSIT 220 (Blum)
15
MAC Bridges
• Like a repeater, a bridge (a.k.a. MAC bridge)
connects segments of a LAN, but a bridge is more
“intelligent.”
• Under “steady-state” conditions, a bridge only
passes a packet from segment A to segment B if the
packet’s destination is on segment B or beyond.
• In other words, while a repeater works at the
physical layer and sees the transmission only as a
wave, a bridge operates at the data-link layer and
understands the transmission as data, in particular its
destination and source addresses.
CSIT 220 (Blum)
16
Bridge
• A bridge is a “computer” with more than one NIC
card operating in promiscuous mode.
– It will probably not be used for anything but this
purpose as the processor will be quite occupied with
this task.
• Each NIC card is attached to a LAN segment.
• All transmissions on these segments are read by
the bridge.
CSIT 220 (Blum)
17
Bridge (Cont.)
• Most bridges are learning or adaptive bridges.
• When such a bridge is first connected, it does not
know which computers are on which segment.
• When a packet first arrives, the bridge “knows”
from which of its cards the message entered (i.e.
what segment the message came from) and it also
reads the source address.
• It has then learned the MAC address of a
computer (the source) and which of its ports that
computer is on.
CSIT 220 (Blum)
18
Bridges (Cont.)
• It does not yet know the side of the
destination computer, so it must transmit
(forward) the packet to all other ports.
• Later when a packet arrives having as a
destination the previous packet’s source
address, the bridge knows whether the
packets must be forwarded or not.
CSIT 220 (Blum)
19
Bridge (Cont.)
• The bridge develops a table of MAC addresses,
and after a time reaches its “steady state” in which
it knows the addresses of most of the active
computers.
– Tables are refreshed periodically in case a computer is
moved.
• It only transmits packets
– That are broadcast
– That are multicast
– That are unicast and have source and destination on
different ports of the bridge.
CSIT 220 (Blum)
20
Repeater vs. Bridge
• Repeaters lead to identical traffic on the connected
segments.
• Bridges reduce the amount of traffic on the
segments, freeing up the transmission line for
increased traffic provided a reasonable amount of
the traffic is intra-segmental (within a segment).
– Increases throughput
• When designing a multi-segmented network, one
wants to maximize intra-segmental
communication.
CSIT 220 (Blum)
21
Bridge
• Recall that error-checking takes place at the
data-link layer.
• Consequently, a bridge transmits only
packets thought to be error free.
• Collisions, noise, interference are not
transmitted across a bridge.
– Repeaters do transmit error ridden packets.
CSIT 220 (Blum)
22
Bridge vs Router
• A restriction on a bridge is that the
connected segments utilize the same
protocol.
• A router serves a somewhat similar purpose
but acts at a higher level (the network layer)
and is “more intelligent.”
• Occasionally a bridge and router are
combined in a product called a brouter.
CSIT 220 (Blum)
23
Long-distance bridge
• One can combine the ideas of the fiber
extension and the bridge to achieve a bridge
that extends over a long distance.
• The computer connected to the first LAN
(segment) is given a second NIC card so
that it can serve as a bridge to a second
LAN (segment).
CSIT 220 (Blum)
24
Long-distance bridge (Fig. 11.6)
CSIT 220 (Blum)
25
Even Greater Distances (Fig. 11.7)
2 half bridges
CSIT 220 (Blum)
26
Bridged Networks
• A bridged network does not necessarily just
form a long line with the end of one
segment bridged to the end of another
segment and a given segment bridged to at
most two others.
B
CSIT 220 (Blum)
B
B
27
Bridged Networks (Cont.)
CSIT 220 (Blum)
28
Possible Problem
Segment X
Bridge 1
Bridge 2
Segment Y
CSIT 220 (Blum)
29
Possible Problem (Cont.)
• If a node on Segment X unicasts a message to a
node on Segment Y, then (in steady state) Bridge 1
will forward it to Segment Y as will Bridge 2. It
will arrive twice.
• If a node on Segment X broadcasts a message,
then Bridge 1 will forward it to Segment Y, then
the message will reach Bridge 2 and be forwarded
to Segment X, where it will reach Bridge 1 and be
forwarded to Segment Y, where ….
• We have an infinite loop (actually two counter
rotating infinite loops).
CSIT 220 (Blum)
30
Logical vs. Physical again
• Physically, loops are good because they can
provide a backup route should one route
fail.
• Logically, loops are bad, they lead to an
infinite cycling of messages.
• So long as the network is logically loop-less
(that is, a tree), it is OK.
CSIT 220 (Blum)
31
STP
• Spanning Tree Protocol (part of the IEEE 802.1
standard) allows for a bridged network that has
physical loops but is logically a tree.
• STP puts the bridges that lead to a loop into a
standby or blocked state (forming a logical tree).
• However, it stores alternate logical trees “in the
event that one bridge is unable to perform its
duties”
– Path redundancy
CSIT 220 (Blum)
32
STP (Cont.)
• Of all the logical trees, one wants the best, that is
the cheapest.
• There will be some “cost function” which will
depend on the throughput of the various
connections, the typical traffic patterns on those
connections and so on.
• STP will select the “minimal” tree, but that could
change, which is another reason that a blocked
bridge may later be activated.
CSIT 220 (Blum)
33
May the circle be unbroken
• Similar to the way one can have more than one
logical tree in case a bridge goes out, one can have
various logical rings in case a connection within
the ring is broken.
• If the network’s physical topology is a ring and it
is broken, then it goes down.
• However, if the network’s physical topology is a
star but it’s logical topology is a ring, then a new
logical ring can be formed should a connection or
node go down.
CSIT 220 (Blum)
34
FDDI Hub
• FDDI Hub contains the electronic circuitry
necessary to detect a broken link and
reconfigure the network.
• The FDDI logical network topology is a
ring, but the physical topology is star.
CSIT 220 (Blum)
35
Switch
• A switch is an intelligent hub.
• The hub operates at the physical layer,
forwarding an incoming signal to all other ports.
• A switch operates at the data-link layer,
forwarding a (unicast) message only to the
designated port.
• A switch is like a many-ported bridge with only
one computer on each segment.
CSIT 220 (Blum)
36
Switch vs. Hub
• For a given message a hub should be faster.
• With increased traffic, switches should
improve throughput, like bridges different
signals can be simultaneously transmitted
on the various segments.
– It allows different messages to be transmitted
“in parallel” — a given message is still sent
serially.
• Hubs are cheaper.
CSIT 220 (Blum)
37
Switch vs. Router
• Routers operate at a higher layer (the
network layer), so they are:
– more “intelligent” improving throughput for
heavier traffic
– slower for handling a given packet
– More expensive
• There’s an intermediate device known as an
“IP switch”
CSIT 220 (Blum)
38
Other references
• http://www.whatis.com
• http://www.webopedia.com
CSIT 220 (Blum)
39