1 scheduling approaches to mac about random access: –simple and easy to implement –in...

1

Scheduling approaches to MAC

• About random access:– Simple and easy to implement

– In low-traffic, packet transfer has low-delay

– However, limited throughput and in heavier traffic, packet delay has no bound. A station of bad luck may never have a chance to transfer its packet.

• Scheduling approach:– provides orderly access to shared medium so that every

station has chance to transfer

2

Scheduling approach—reservation protocol• The time line has two kinds of periods:

– Reservation interval of fixed time length

– Data transmission period of variable frames.

• Suppose there are M stations, then the reservation interval has M minislots, and each station has one minislot.

• Whenever a station wants to transfer a frame, it waits for reservation interval and broadcasts reservation bit in its minislot.

• By listening to the reservation interval, every station knows which stations will transfer frames, and in which order.

• The stations having reserved for their frame transfer their frames in that order

• After data transmission period, next reservation interval begins.

3

Reservation protocol

10 1 2 3 4 5 6 7

1 1 1 3 50 1 2 3 4 5 6 7

1 1 3 7

Reservation interval

Frames-transmission

4

Scheduling approach—polling protocol

• Stations take turns accessing the medium:– At any time, only one station has access right to

transfer into medium– After this station has done its transmission, the access

right is handed over (by some mechanism) to the next station.

– If the next station has frame to transfer, it transfers the frame, otherwise, the access right is handed over to the next next station.

– After all stations are polled, next round polling from the station 1 begins.

5

Centralized polling vs. distributed polling

• Centralized polling: a center host which polls the stations one by one

• Distributed polling: station 1 will have the access right first, then station 1 passes the access right to the next station, which will passes the access right to the next next station, …

6

t1 32 4 5 1 2

polling messages

packet transmissions

… M

Figure 6.28

Interaction of polling messages and transmissions in polling systems

7

Figure 6.30

Token-passing rings – a distributed polling network

Station interfaces: are connected to form a

ring by point-to-point lines Stations: are attached to the ring by station interfaces.

Note: point-to-point lines, not a shared bus.

Station interfaces have important functions.

Token: a small frame, runs around the ring, whichever gets the token, it has the right to transmit data frames.The information flows in one direction.

token

8

listen mode

1 bit delay

transmit mode

delay

to station from station

inputfromring

outputtoring

Figure 6.30

Token-passing rings – a distributed polling system

Two modes of interface: 1. Listen mode, like a repeater but with some delay, because every arriving bit will be copied into a 1-bit buffer and then copied out. While in the buffer, the bit can be monitored, i.e., inspected and even modified. Therefore, 1-bit delay 2. Transmit mode: transmit frames from its attached stationWhen monitoring, if it finds the passingbits are a data frame with its attached stationas destination, then it catches the frame,i.e., copy the bits into its station (mayrefrain from outputting them).

If it finds the passing bits are free token, then if the station has information to send, it will change the token to busy (set one bit to 1) and change to transmit mode.

to station from station

9

Removing a frame

A transmitted frame needs to be removed (absorbed).Who removes it?

Choice one: destination stationChoice two: transmitting station itself.

Choice two is preferred because the destination can insert ACKinto the frame and transmitter will get the ACK from its owntransmitted frame.

10

dd

ddd dd

dd

d

d

dd

d

d

a) b) c)

Busy token Free token

Figure 6.31

Approaches to token reinsertion: a). multitoken, b). single token, c)single packet

a). The free token is inserted immediately after the last bit of data frame is sentb). Insert the free token after the last bit of busy token is receivedc). Insert the free token after the last bit of data frame is received.

Generally: when a station seizes the token, it can only transmit limited number of frames or limited time interval to avoid occupying too long.

11

Comparison of scheduling & random access• Scheduling

– Methodical orderly access: dynamic form of time division multiplexing, round-robin (only) when the stations have information to send.

– Less variability in delay, supporting applications with stringent delay requirement. In high load, performance is good. E.g., token-ring may reach nearly 100 percent of performance when all stations have plenty of information to send.

– Some channel bandwidth carries explicit scheduling information

• Random access– Chaotic, unordered access– If rich bandwidth and light load, random access has low delay,

otherwise, delay is undeterminably large.– Quite a lot bandwidth is used in collision to alert stations of the

presence of other transmissions.

12

Channelization

• Suppose M stations generate steady flow of information

• Divide shared medium into M channels so that each station is allocated one channel

• Four channelization mechanisms:– FDMA– TDMA– WDMA– CDMA

13

A1 A2

B1 B2

C2C1

A2B1 B2 C2C1

(a)

(b)A1

Shared Line

Dedicated Lines

Figure 5.43

Time-Division Multiplexing (TDM)

14

2

24

1

MUXMUX

1

2

24

24 b1 2 . . .b2322

frame

24 . . .

. . .

Figure 4.4

T-1 carrier system uses TDM to carry 24 digital signals in telephone system

15

A CBf

Cf

Bf

Af

W

W

W

0

0

0

(a) Individual signals occupy W Hz

(b) Combined signal fits into channel bandwidth

Figure 4.2

Frequency-Division Multiplexing (FDM)

16

MUX deMUX1, 2, n

Figure 4.1

1

2

n

Optical fiber

1

2

n

Wavelength-division multiplexing (WDM)

17

CDMA (Code-division multiple access)• Transmissions from different stations occupy

entire frequency band at the same time• Different codes are used to separate different

transmissions• A code is a unique binary pseudorandom sequence

(such as c1,c2,…,cG) (each ci is a +1 or –1 signal)• A binary bit b (i.e., +1 signal) is spread by

multiplying it with the code at the sender (which is a sequence of binary information)

• At the receiver, the received binary string is multiplied by the same code to get back original bit.

18

CDMA (cont.)• Since:

– b(c1+c2+…+cG) at the sender– The received information is multiplied by c1,c2,…,cG to get

b(c12+c2

2+…+cG2) = b1=b.

– Due to each ci2=1 no matter ci=1 or –1

– So the receiver gets back b.

• However, if a different code d1,d2,…,dG is used:– The multiplication will be c1 d1+c2 d2

+…+cG dG

– Since ci and di are independent, there will be equal number of +1 and –1, so the result will be 0,

– Thus the b can not be obtained.

19

CDMA (cont.)

• In order for different codes are randomized, the sequence length G must be long enough.

• G is called spreading factor.

• Using LSFR (Left-Shift-Feedback-Register).

21

Binaryinformation

R1 bpsW1 Hz

Unique user binary random

sequence

Digitalmodulation

Radio antenna

Transmitter from one user

R >> R1bpsW >> W1 Hz

CDMA Spread Spectrum Signal

• User information mapped into: +1 or -1 for T sec.• Multiply user information by pseudorandom binary pattern of G “chips” of +1’s and -1’s • Resulting spread spectrum signal occupies G times more bandwidth: W = GW1

• Modulate the spread signal by sinusoid at appropriate fc

22

Signal and residualinterference

Correlate touser binary

random sequence

Signalsfrom all

transmittersDigital

demodulation

Binaryinformation

CDMA Demodulation

• Recover spread spectrum signal• Synchronize to and multiply spread signal by same pseudo-random

binary pattern used at the transmitter• In absence of other transmitters & noise, we should recover the original

+1 or -1 of user information• Other transmitters using different codes appear as residual noise

23

R0 R1 R2

g(x) = x3 + x2 + 1

g0g2 g3

The coefficients of a primitive generator polynomialdetermine the feedback taps

Time R0 R1 R2

0 1 0 01 0 1 02 1 0 13 1 1 0 4 1 1 1 5 0 1 1 6 0 0 1 7 1 0 0

Sequence repeatsfrom here onwards

output

Pseudorandom pattern generator• Feedback shift register with appropriate feedback taps

can be used to generate pseudorandom sequence

24

Channelization in Code Space• Each channel uses a different pseudorandom code• Codes should have low cross-correlation

– If they differ in approximately half the bits the correlation between codes is close to zero and the effect at the output of each other’s receiver is small

• As number of users increases, effect of other users on a given receiver increases as additive noise

• CDMA has gradual increase in BER due to noise as number of users is increased

• Interference between channels can be eliminated if codes are selected so they are orthogonal and if receivers and transmitters are synchronized– Shown in next example

25

Example: CDMA with 3 users• Assume three users share same medium• Users are synchronized & use different 4-bit orthogonal codes:

{-1,-1,-1,-1}, {-1, +1,-1,+1}, {-1,-1,+1,+1}, {-1,+1,+1,-1},

+1 -1 +1

User 1 x

-1 -1 +1

User 2 x

User 3 x

+1 +1 -1 SharedMedium

+

Receiver

26

Channel 1: 110 -> +1+1-1 -> (-1,-1,-1,-1),(-1,-1,-1,-1),(+1,+1,+1,+1)Channel 2: 010 -> -1+1-1 -> (+1,-1,+1,-1),(-1,+1,-1,+1),(+1,-1,+1,-1)Channel 3: 001 -> -1-1+1 -> (+1,+1,-1,-1),(+1,+1,-1,-1),(-1,-1,+1,+1)Sum Signal: (+1,-1,-1,-3),(-1,+1,-3,-1),(+1,-1,+3,+1)

Channel 1

Channel 2

Channel 3

Sum Signal

Sum signal is input to receiver

27

Example: Receiver for Station 2• Each receiver takes sum signal and integrates by code

sequence of desired transmitter• Integrate over T seconds to smooth out noise

x

SharedMedium

+

Decoding signal from station 2

Integrate over T sec

28

Sum Signal: (+1,-1,-1,-3),(-1,+1,-3,-1),(+1,-1,+3,+1) Channel 2 Sequence: (-1,+1,-1,+1),(-1,+1,-1,+1),(-1,+1,-1,+1)Correlator Output: (-1,-1,+1,-3),(+1,+1,+3,-1),(-1,-1,-3,+1)Integrated Output: -4, +4, -4Binary Output: 0, 1, 0

Sum Signal

Channel 2Sequence

CorrelatorOutput

IntegratorOutput

-4

+4

-4

Decoding at Receiver 2

X

=

29

W1= 0 W2=0 00 1

W4= 0 00 1

0 00 1

0 00 11 11 0

W8=

0 00 1

0 00 1

0 00 11 11 0

0 00 1

0 00 1

0 00 11 11 0

0 00 1

0 00 1

0 00 11 11 0

1 11 0

1 11 0

1 11 00 00 1

Walsh Functions• Walsh functions are provide orthogonal code sequences by mapping 0

to -1 and 1 to +1: each row is a code, rows are orthogonal. • Walsh matrices constructed recursively as follows:

W2n=Wn Wn

Wn Wnc

30

LAN bridges

• Repeater: used to connect two or more networks at physical layer

• Bridge: used to connected two or more networks at data link or MAC layer

• Router: used to connect two or more networks at network layer

• Gateway: connect two or more networks at higher layers.

31

LAN bridges’ functions• Bridges are generally used to connect LANs by

– Extending a LAN which reach saturation, called bridged/extended LAN

– Connect different departments’ LANs, these LANs

• Use different network layer protocols, bridges are fine because bridges operate at data link layer which supports different network layer protocols

• Located in different building, bridges are fine because bridges can be connected by point-to-point link

• Are different LANs: bridges need to have ability to convert between different frame format

– Security is big problem in LAN, why?

• So bridges need to have some ability of dealing with security issue such as filtering frames, controlling flow of frames in and out.

– Bridge is better than repeater when connect two exact same LANs because bridge has the ability to reduce traffic by confining local traffic.

– Bridge will monitor MAC address of frames so it can not be in physical layer

– Bridge have no routing ability so it is not in network layer.

32

Bridge

Network

Physical

Network

LLC

PhysicalPhysicalPhysical

LLC

MAC MACMAC MAC

Figure 6.80

Bridges generally connect the same types of LANs, so they generally

operate at MAC layer.

Interconnection by a bridge

33

Bridge

S1 S2

S4

S3

S5 S6

LAN1

LAN2

Figure 6.79

A bridged LAN

Port #1

Port #2

34

Three type bridges• Transparent bridges: means that stations are completely

unaware of the presence of bridges. Used in Ethernet LANs. No burden in stations, bridges take care of all connection related functions.

• Source routing bridge: used in token-ring and FDDI LANs. Burden on stations. Source station needs to give route to the destination. Bridges just forward frames based on the route in the frame.

•Mixed-media bridges: used to interconnect LANs of different types. These bridges have abilities of converting between LANs.

35

Transparent bridges

• Three basic functions:– Forwards frames from one LAN to another– Learns where stations are attached to the LAN– Prevents loops in the topology

• A transparent bridge is configured in “promiscuous” mode

36

Forwarding and forwarding table of bridges

MAC address Port

1. Port is a physical interface, a bridge have two or more ports2. For every station in bridged LAN, the port number indicates which part (direction) of the bridge this station is attached to.3. When a frame comes to a bridge, the bridge will forward to the port corresponding to the MAC address in the table, which is the destination physical address in the frame.4. Question: how to establish forwarding table?

Manually set up by administrator. Good? No, automatically set up by self learning

37

Bridge learning

When a bridge receives a frame, it searches through the forwarding table:

1. For source address, if not found, adds source address along coming port # into table

2. For destination address, if found, forwards the frame to the corresponding port, except the corresponding port # being same as the coming port #

3. otherwise (not found), floods the frame to all ports except the coming port

38

Bridge1

S1 S2

Bridge 2

S3 S4 S5

Address Port Address Port

port 1 port 2 port 1 port 2

LAN1 LAN2 LAN3

S1 1 S1 1S3 2 S3 1

S4 2S4 2

S2 1

Figure 6.85

1.S1->S5 2.S3->S2 3.S4->S34.S2->S1

Any problem with it?

39

LAN topology is dynamic

Life is never static in real world. The bridged LAN change constantly.

Add station: Easy, learn again.

Move station: When find same MAC address, but coming from a different port, update the port number.

Remove station: Timer, associate every entry (a station) in forwarding table a timer, when timer times out, remove the entry from the table. Moreover, when receiving a frame with a sourceMAC address, if its entry is already in the table, thenrefresh the timer.

Any more problem?

40

Bridge1

S1 S2

Bridge 2

S3 S4 S5

port 1 port 2 port 1 port 2

LAN1 LAN2 LAN3

Figure 6.85

S1->S5

Bridge3

Loop in the topology will cause flooding forever

41

Spanning tree to break loop• Main idea: maintain a spanning tree to

include all stations but disable some ports automatically. Thus, remove loop.

• Assumptions: unique LAN IDs, unique bridge IDs, and unique port IDs. The

lowest ID is used to break a tie.

42

Steps of spanning tree1. Select a root bridge which is the bridge with the lowest

bridge ID.2. For each bridge except the root bridge, determine the root

port which is the port with least-cost path (& lowest ID) to root bridge.

3. For each LAN, select a designated bridge, which is the bridge offering the least-cost path (& lowest ID) from the LAN to root bridge. The port connecting the LAN and the designated bridge is called a designate port.

4. All root ports and designated ports are put into forwarding table. These are only ports that are allowed to forward frames. The other ports are placed into a “blocking” state.

43

LAN1

LAN2

LAN3

B1 B2

B3

B4

B5

LAN4

(1)

(2)

(1)

(1)

(1)

(1)

(2)

(2)

(2)

(2)

(3)

Figure 6.86

44

L A N 1

L A N 2

L A N 3

B 1 B 2

B 3

B 4

B 5

L A N 4

( 1 )

( 2 )

( 1 )

( 1 )

( 1 )

( 1 )

( 2 )

( 2 )

( 2 )

( 2 )

( 3 )

R

R

R

R

D

D

DD

Figure 6.87

45

Source routing bridges• Putting burden on end stations, bridges are mainly

responsible for forwarding.• Each station determines the route to the

destination and put the routing information in the header of the frame.

• Source routing information is inserted only when source and destination are in different LANs and is indicated by I/G bit in source address.

46

RoutingControl

RouteDesignator-1

RouteDesignator-2

RouteDesignator-m

DestinationAddress

SourceAddress

RoutingInformation

Data FCS

2 bytes 2 bytes 2 bytes 2 bytes

Figure 6.88

Frame format for source routing

1. Control field contains type of frame, routing information and length of it.

2. Designator contains a 12-bit LAN number and a 4-bit bridge number

3. The highest bit in source address indicates whether it is source routing.

Question: how to find the route in the first place?

51

Mixed-media bridges• Such as connect Ethernet and token-ring networks• Convert between two address representations• Ethernet has maximum size of 1500 bytes but token-

ring has no explicit limit. Drop frame if it is too long because bridges do not do segmentation

• Token-ring has three status bits A, C and E, but

no these bits in Ethernet, so no conversion for them• Different transmission rate, so bridges need buffer to

store extra frame temporarily.

52

IEEE LAN standards

• 802.2 LLC• 802.3 CSMA-CD (I.e., Ethernet)

– Frame length 64 –1518 (or client data: 46—1500)

• 802.4 token-bus• 802.5 token-ring

– FDDI (Fiber Distributed Data Interface)

• 802.11 wireless networks

53

Ethernet --history• 1970s, Robert Metcalfe etc. Xerox, connecting workstations• 1980s, DEC, Intel, Xerox, “DIX” Ethernet standard 10Mpbs• The basis for IEEE 802.3 standard, “thick” coaxial cable in

1985.• 802.3 and “DIX” Ethernet differ in one header field definition.• Expended to “thin” coaxial, twisted-pair, optical fiber.• 1995: 100Mbps, 1998: 1Gbps, 2002: 10Gbps.

54

Ethernet -- basis

• Bus based coaxial cable

• CDMA-CD with 1-persistent

• Minislot time: two propagation delays

• 10Mbps, 2500 meters, 4 repeaters

• (Truncated) binary exponential backoff algorithm

• Three periods and performance

55

Ethernet --backoff

• (Truncated) binary exponential backoff algorithm– Whenever collision, wait for B minislots

– B is determined as followed:• After 1th collision, B is selected from 0 to 1

• After 2th collision, B is selected from 0 to 3

• After 3th collision, B is selected from 0 to 7

• …

• After nth collision, B is selected from 0 to– 2n-1 if n<10 and 210-1 if n<=16, otherwise, give up.

56

Ethernet--performance• Three periods: idle, contention, transmission

• During saturation, no idle period– A L/R transmission period

– A tprop period for finding the end of transmission

– A contention period of B minislots, in average B=e=2.71.

• So performance is:– (L/R)/(L/R+ tprop+2e tprop)=1/(1+(1+2e) tpropR/L)=

– 1/(1+(1+2e)a)=1/(1+6.44a), where a= tprop R/L.

57

Ethernet--performance

• Why B=e in average?• Suppose n stations and each station transmits during a

contention period with probability p,– The probability that one station transmits successfully is

Psuccess=np(1-p)n-1.– For maximum throughput, p should be selected to maximize

Psuccess, i.e., p=1/n.– So Psuccess

max=n(1/n)(1-1/n)n-1=(1-1/n)n-11/e.

• Since probability of success in one minislot is Psuccessmax ,

the average number of minislots that elapse until a station captures the channel is – 1/ Psuccess

max =e.

58

IEEE 802.3 MAC Frame

Preamble SDDestination

addressSource address

Length Information Pad FCS

7 1 6 6 2 4

64 - 1518 bytesSynch Startframe

802.3 MAC Frame

Every frame transmission begins “from scratch” Preamble helps receivers synchronize their clocks

to transmitter clock 7 bytes of 10101010 generate a square wave Start frame byte changes to 10101011 Receivers look for change in 10 pattern

59





7 1 6 6 2 4


0 Single address

1 Group address

• Destination address• single address• group address• broadcast = 111...111

Addresses• local or global

• Global addresses• first 24 bits assigned to manufacturer;• next 24 bits assigned by manufacturer• Cisco 00-00-0C• 3COM 02-60-8C

0 Local address

1 Global address

802.3 MAC Frame

60





7 1 6 6 2 4


802.3 MAC Frame

Length: # bytes in information field Max frame 1518 bytes, excluding preamble & SD Max information 1500 bytes: 05DC

Pad: ensures min frame of 64 bytes FCS: CCITT-32 CRC, covers addresses, length,

information, pad fields NIC discards frames with improper lengths or failed CRC

61

DIX Ethernet II Frame Structure

• DIX: Digital, Intel, Xerox joint Ethernet specification• Type Field: to identify protocol of PDU in information field, e.g. IP, ARP• Framing: How does receiver know frame length?

– physical layer signal, byte count, FCS



Type Information FCS

7 1 6 6 2 4


Ethernet frame

62

IEEE 802.3 Physical Layer

(a) transceivers (b)

10base5 10base2 10baseT 10baseFX

Medium Thick coax Thin coax Twisted pair Optical fiber

Max. Segment Length 500 m 200 m 100 m 2 km

Topology Bus Bus StarPoint-to-point link

Table 6.2 IEEE 802.3 10 Mbps medium alternatives

Thick Coax: Stiff, hard to work with T connectors flaky

Hubs & Switches!

63

Ethernet Hubs & Switches

(a)

Single collision domain

(b)

High-Speed backplane or interconnection fabric

Twisted Pair CheapEasy to work withReliableStar-topology CSMA-CD

Twisted Pair CheapBridging increases scalabilitySeparate collision domainsFull duplex operation

64

CSMA-CD

0

5

10

15

20

25

30

0

0.06

0.12

0.18

0.24 0.3

0.36

0.42

0.48

0.54 0.6

0.66

0.72

0.78

0.84 0.9

0.96

Load

Avg

. T

ran

sfe

r D

ela

y

a = .01a = .1a = .2

Ethernet Scalability

• CSMA-CD maximum throughput depends on normalized delay-bandwidth product a=tprop/X=tpropR/L

• x10 increase in bit rate = x10 decrease in X• To keep a constant need to either: decrease tprop (distance) by x10; or

increase frame length x10

65

Fast Ethernet

• IEEE 802.3u• 100Mbps• CSMA-CD is sensitive to a, since rate increases

10 times, to keep same performance with Ethernet, ???

Either increase the frame size 10 times (to 640 bytes) or reduce maximum Distance 10 times (250 meters).

The final decision is to keep the frame size and procedure unchanged butTo define a set of physical layers based on hub (start) topology involving Twisted-pair and optical fiber.

66

Fast Ethernet

• Two modes:– All incoming lines are logically connected to a

single collision domain and CSMA-CD is used– Incoming frames are buffered and then switched

internally within the hub, CSMA-CD is not used but use multiplexing and switching.

• Fast Ethernet is deployed in departmental networks– Aggregate traffic from shared 10Mbps LANs– Provide greater bandwidth to a server– Provide greater bandwidth to certain users.

67

Gigabit Ethernet

• 802.3z, 1998, 1Gbps.

• Frame size extended to 512 bytes.

• Moreover frame bursting: a burst of small frames.

• Preserve the frame structure but operate primarily in a switched mode.

68

Fast Ethernet

100baseT4 100baseT 100baseFX

MediumTwisted pair category 3

UTP 4 pairs

Twisted pair category 5

UTP two pairs

Optical fiber multimode

Two strands

Max. Segment Length

100 m 100 m 2 km

Topology Star Star Star

Table 6.4 IEEE 802.3 100 Mbps Ethernet medium alternatives

To preserve compatibility with 10 Mbps Ethernet:• Same frame format, same interfaces, same protocols• Hub topology only with twisted pair & fiber• Bus topology & coaxial cable abandoned• Category 3 twisted pair (ordinary telephone grade) requires 4

pairs• Category 5 twisted pair requires 2 pairs (most popular)• Most prevalent LAN today

69

Gigabit EthernetTable 6.3 IEEE 802.3 1 Gbps Fast Ethernet medium alternatives

1000baseSX 1000baseLX 1000baseCX 1000baseT

Medium

Optical fiber

multimode

Two strands

Optical fiber

single mode

Two strands

Shielded copper cable

Twisted pair category 5

UTP

Max. Segment Length

550 m 5 km 25 m 100 m

Topology Star Star Star Star

• Slot time increased to 512 bytes• Small frames need to be extended to 512 B• Frame bursting to allow stations to transmit burst of short frames• Frame structure preserved but CSMA-CD essentially abandoned• Extensive deployment in backbone of enterprise data networks

and in server farms

70

10 Gigabit EthernetTable 6.5 IEEE 802.3 10 Gbps Ethernet medium alternatives

10GbaseSR 10GBaseLR 10GbaseEW 10GbaseLX4

Medium

Two optical fibersMultimode at 850 nm

64B66B code

Two optical fibers

Single-mode at 1310 nm

64B66B

Two optical fibers

Single-mode at 1550 nmSONET compatibility

Two optical fibers multimode/single-mode with four wavelengths at 1310 nm band8B10B code

Max. Segment Length

300 m 10 km 40 km 300 m – 10 km• Frame structure preserved• CSMA-CD protocol officially abandoned• LAN PHY for local network applications• WAN PHY for wide area interconnection using SONET OC-

192c • Extensive deployment in metro networks anticipated

71

Server

100 Mbps links

10 Mbps links

ServerServer

Server

100 Mbps links

10 Mbps links

Server

100 Mbps links

10 Mbps links

Server

Gigabit Ethernet links

Gigabit Ethernet links

Server farm

Department A Department B Department C

Hub Hub Hub

Ethernet switch

Ethernet switch

Ethernet switch

Switch/router Switch/router

Typical Ethernet Deployment

72

IEEE 802.5 Ring LAN

• Unidirectional ring network• 4 Mbps and 16 Mbps on twisted pair

– Differential Manchester line coding

• Token passing protocol provides access Fairness Access priorities Breaks in ring bring entire network down

• Reliability by using star topology

73

Wiring Center

A

B

CD

E

Star Topology Ring LAN• Stations connected in star fashion to wiring closet

– Use existing telephone wiring• Ring implemented inside equipment box• Relays can bypass failed links or stations

74

Token frame format

SD FCAC Destinationaddress

Source address

Information FCS

1 4

ED

6 61 11

FS

1

Data frame format

Token Frame Format

SD AC ED

P P P T M R R RAccess control

PPP=priority; T=token bitM=monitor bit; RRR=reservationT=0 token; T=1 data

Starting delimiter

J, K nondata symbols (line code)J begins as “0” but no transitionK begins as “1” but no transition

0 0J K 0 J K 0

Ending delimiter

I = intermediate-frame bitE = error-detection bitI EJ K 1 J K 1

75

Frame control

FF = frame type; FF=01 data frameFF=00 MAC control frameZZZZZZ type of MAC control

F F Z Z Z Z Z Z

Framestatus

A = address-recognized bitxx = undefinedC = frame-copied bit

A C x x A C x x

SD FCAC Destinationaddress

Source address

Information FCS

1 4

ED

6 61 11

FS

1

Data frame formatData Frame Format

Addressing 48 bit format as in 802.3

Information Length limited by allowable token holding time

FCS CCITT-32 CRC

76

Other Ring Functions• Priority Operation

– PPP provides 8 levels of priority– Stations wait for token of equal or lower priority– Use RRR bits to “bid up” priority of next token

• Ring Maintenance– Sending station must remove its frames– Error conditions

• Orphan frames, disappeared token, frame corruption• Active monitor: a station responsible for error corrections.

– MAC control frames and protocol are needed. • Beacon frames for detecting neighbors and broken links.• Election protocol.

77

Ring Latency & Ring Reinsertion• M stations• b bit delay at each station

– b=2.5 bits (using Manchester coding)

• Ring Latency: ’ = d/ + Mb/R seconds ’R = dR/ + Mb bits

• Example– Case 1: R=4 Mbps, M=20, 100 meter separation

• Latency = 20x100x4x106/(2x108)+20x2.5=90 bits– Case 2: R=16 Mbps, M=80

• Latency = 840 bits

A A A

A A A A

t = 0, A begins frame t = 90, returnof first bit

t = 400, transmission of

last bit

A

t = 490, last bit received, reinsert

token

t = 0, A begins frame t = 400, transmitlast bit

t = 840, arrivalfirst frame bit

t = 1240, last bit received, reinsert

token

(b) High Latency (840 bit) Ring

(a) Low Latency (90 bit) Ring Efficiency=400/490=82%

Efficiency=400/1240=32%

results in low efficiency.Reinsert token after frame return (400: frame length)

A A A

A A A A

t = 0, A begins frame t = 90, returnof first bit

t = 210, return of header

A

t = 400, last bit enters ring, reinsert token

t = 0, A begins frame t = 400, transmitlast bit

t = 840, arrivalfirst frame bit

t = 960, last bit of header received, reinsert token

(b) High Latency (840 bit) Ring

(a) Low Latency (90 bit) Ring (suppose header=120 bits) Efficiency=400/400=100%

Efficiency=400/960=42%

Reinsert token immediately after completion of frame transmission: no idle period, 100%.

80

Fiber Distributed Data Interface (FDDI)• Token ring protocol for LAN/MAN• Counter-rotating dual ring topology• 100 Mbps on optical fiber• Up to 200 km diameter, up to 500 stations• Station has 10-bit “elastic” buffer to absorb timing

differences between input & output• Max frame 40,000 bits• 500 stations @ 200 km gives ring latency of

105,000 bits• FDDI has option to operate in multitoken mode

81

A

E

DC

B

X

Dual ring becomes a single ring

82

SD DestinationAddress

Source Address

Information FCS

8 4

EDFC

6 61 11

FS

1

PRE

Preamble

SD FC EDToken Frame Format PRE

Frame control

Data Frame Format

CLFFZZZZ C = synch/asynch L = address length (16 or 48 bits)FF = LLC/MAC control/reserved frame type

CLFFZZZZ = 10000000 or 11000000 denotes token frame

FDDI Frame Format

1 scheduling approaches to mac about random access: –simple and easy to implement –in...

Documents