r .....

aoN
,, coNGP
•
••
............
, Congestion
. is an imPortunt iss
. .
························
·········-··············· . . pocket switched
............ .....
. computer
. ············ •••••
. uc that can anse 111
networks
.
m
.
which the
k .
perfo
11 ctWork ,,
t111n 11 •
·
N!iTWOftK LAYER

).
~ 9.55

If suddenly , a stream of packet start arriving on three or four input lines

und all need the same output line (see fig. 9.26). In this case, a queue will be
• · 8 s1ruation Ill cts the b cc built up.
Congesuon is the presence of too many p~c 111 Of
rk . degraded due to su nc1 I
netwo is k iay occur when the load on the network (I , ·
• 1·n a networ n Ilic
Congcsuon , t the network) is greater than the capac'tI y ,c, of
Input line 1
be Of packets s,nt o
(8 )
num r be fpackets a network can handle.) the output line
Input fine 2
network (i.e. the num r o . .
cept of congesuon m a subnet.
• Fieure 9.25 shows t econ h . Input fine 3
- . fi 9 25 when the number of packets dumped mto th~ sub 11etb Aout,;ir
• Asshownm 1g. • . h II d .
. 'thi n I•ts C°'Tving capacity, t ey are a e11vered. At th·IS Sta Y
the hosts IS WI -·, .
delivered is proportional to the number of packets senr and &e
number of pac kets
. Fig. 9.26 Data from three input rines at same time
no congestion take place.
,_rt: c increases too far, the routers are no • longer able to cope
• As the ul1.1111 • • up and If there is insufficient me.norj to hold all the packets, the packet will be
they start losing packets \\1th further mcrease m the traffic, perfonnan
. ~ Jost.
degrades more and more, packets are Iost and congest10n worsens.
Increasing the memory to unlimited size does not solve the problem. This is
because, by the time packets reach front of the queue, they have already
timed out (as they waited in the queue). When timer goes off source
I transmits duplicate packet that are also added to the queue. Thus same
I
I packets are added again and again, increasing the load all the way to the
I destination.
I 2. Congestion in a subnet can occur if the processors are slow. Slow speed
CPU at routers will perfonn the rou:ine tasks such as queuing buffers,
I
I
updating table etc slowly. As a result of this, queues are built up even though
there is excess line capacity.
I

3. Congestion is also caused by slow liDks. This problem will be solved when
l. high speed links arc u.,cd. But it is not always the case. Sometimes increase
bl link bandwidth can f\lrtbcr deteriorate the congestion problem as higher
I
I Padfeta 18111
lpllll links IM)' Diab Ibo netwk IDOII unbl)NM'A"Cl,
Fig. 1.25 Concapt" 4. Oliflf SM• - - - ...... .. If a roum' doel not bave me buffers, it
. traffic, J)Cffonnance . . . _ • lli#fi,.wl . . • e. • newly 111m111.--. When these packets are
fig. f.16) 'fi onK tat . . 111111 • 1t tJilla .a.1he 1imer goes off. Such
- I • '« a l>tl"t ..... . apia and epin until the some gets
Conpe~D • II► inm of
'IJ! j . 1111 11 H of'IIIII ,.._ 1'helob e multipk
in .... 11'1 : - - wlll bao tb, OQllllniOD to take place at the senclifll..,.
 tJoJt conrro1 . I;\;\
9 . 5◄ 9,9.2 conl'' ,.
. trul rc,a~ 111
icchni~ucs ond 111cc ha111 s11111 11,111 .,

,., ('on~~fi11n on .
eonic~uon,
(
. t,cforc 11hupf)( '· .
.
.
l .
.11 or remove co11gcst1<111, nl er ,1 hus h1111
·hani~111s urc divided rnto two c111c~ori ,
c,111 c' I
Pc11cd
11 lcr
,
Prt
i,
NUW<HtK lA'r'rR
9.57

• The re1run11mi1111ion policy and lhe rc1rn11,mi~~ii,n timcr11 n«d to be dc!liiµx:d

. con1rol mrc ., I I c~, o 11c Ci
Ctin~1•.111on . from happening anu t 1c 01 1cr CLllcgury re, to op1l111l1.c cflicicncy and at the •amc time prevent the e<mge,ti,,n.
p,r1·rn IS
die congcsuon llovc~ 11t,.
· Coll .,~ 2. Window PolJcy
alltr i1 h,'IS i;ikcn place. kt111,,
., • To implement window policy, 1elcctive reject window method is u.,ed for
Co111•01uon congeHtion control.
r.01111111
• Selecti ve Reject method is preferred over Go-hack-n window as in Go-b-c1Ck-n
method, when limcr for a packet times out, several p-.ac~ are resent,
although some may have arrived safely at the receiver. Thllll, this duplication
may make conge11tion worse.
• Selective reject method sends only the specific lo~ or damaged pack<.:15.
3, Acknowledgem ent Policy
The acknowledgement policy imposed by the receiver may also affect
congestion.
• If the receiver does not acknowledge every packet it receives it may slow
down the sender and help prevent congestion.
• Acknowledgments also add to the traffic load on the network. Thus, by
Fig. 9.27 Types of Congestion Control Methods sending fewer acknowledgements we can reduce load on the network.
I 'J

These 1wo categories are (see fig. 9.27): • To implement it, several approaches can be used :
I. Open loop I. A receiver may send an acknowledgement only if it has a packet to be
2. Clcsed loop sent.
2. A receiver may send an acknowledgement when a timer expires.
Open Loop Congestion Control
3. A m:eiver may also decide to ICbowleclp only N packets at a time.
• In thi, method, policies are used to

• • Congestion control iJ bandied eitherprevent the conoestion

The various ......._.,_
o
by the source or by the
t..J. it
1,11;;10re hilp,-
destfnada4.
4-~Poll oJ
• A router JMY. discard 1w Nlllitive ,.._ WMII ll c111n6n ia likely to

0• RetraumJqfo
nPoUq
• The aender retrarunut,
"""'IUUI used for open loop ,
consest1on control are .
· •
blppca. ,___-
«•• da •••,_.time
---
Such a clilolrdlllapoky-, . . -
- ·· .
may
nota1111....,1Nr-1R tpfdn

3 -
corrupted. t" • Jlll4d, if ufeel, that the ,Packet it has ICJU •
•&
"'""• ..._.. .
--•IIOeaa.lin
~
~
S, Admk:tfta .....

..,,. k • - lllld ro ~ 800d'a·~sm1ss1on

~8 1
increase. t~e congestion in the t'. •
• All . . . . . . . . . . . . . . lli• . . . . .
plMllt . . . . . . . . . . ., . . .. .
policy to prevenl
• Swltchee In a flow ftnt chect~ _ , . ....
admitting it to the network.
/ . ,jrtual circuit connection if ..,
8 ' ,._ 111 NETWORK LAYER
9.58 d ny esta bl'sh111S
1 . 8 Possibility of 1utu,-e congestio Cre 1•1
!IJl.lfer can e rwork or if there is n.
• A . in(h(ne 2 • Choke Packet d t r node sends a spec1th
·a1
C(l/lfeslJOl'I ·o.o contro1 is method of congestion control. congeste rou er O .
colllestl hanisms try to remove the congestio • In Ill
type of packet ca lled choke packet to the source to I orm
·w 1t about C
. n contr0I mec n aOcr \
CJosed LOOP congesoo
• Closed IOOP . f congestion. .
s sed loop congesuon control are : does not · r
m,orm ·
its upstream no de about the coogesllon
it h3Ppen · !hods used for clo • Herc, congested node
• Jbe ,-arioUS me as in backpressure method.
• In choke packet method, congested node sends a wammg rec
· di tly to the
...,,essure estion control that starts with a nod
1 BaC•r od -to-node cong e ¾d source station i.e. the intermediate nodes through which the packet has
• sure is an e . . of data flow, to the source as sho
• 13ackpres . the opposite du-ec□on WI! in traveled are not warned (see fig, 9.29)
propagates. JD
figure 9-28
Choke Packet

Data flow
t¢team t I
Fig. 9.29 Choke Packet Method
F',g. 928 BackpreSSure method
3. Implicit Signaling
• Th backpressure technique can be applied only to virtual circuit networks. In • In implicit signaling, there is no communication between the congested node
sue~ virtual circuit each node bows the upstream IK>"..e fiom which a dala or nodes and the source.
flow is coming. • The source guesses that there is a congestion somewhere in the network when
• In this method of congestion contro~ the ~ node stops ~ - it does not receive any acknowledgment. Therefore die delay in receiving an
from the immediate upstream node or nodet. acknowledsment is interpreted• cow.«tliub Ja the network.
• This may cause the upstream node Oil aoda to become ~lad • Onsemine du, con,eltioa.1M --••cloWII.
in tum, reject data 1i'om dieir gp.mn dot aodel. • Tbf, 1JJJe of CCQ1111ioaCIGldlOl...,itlllld'li,l'CP.
• As shown in fig. 9.28 node 3 is eoll,ated and it stops receiving ~ 4,5 .,. . . . .,,,.
infonns its ups:ream node 2 to slow down. Node 2 in turns, may be • fa tlaftmediad. tbf TT ••tel. . NIii ltj t;ftt .......... lOl!nle or
and infonns node I to slow down. Now aode 1 may create a conge,don ad
.imm.s die --.e 111111, ro n down. In thi, way the conpsti911 it
dlltfnadoa.10 IDfoa uouttbo oona•••
• • Bxplldt_,,. . . . . . . . . . . . :1.11111' idt- pmd
on node 3 i, moved backward to the source to ; ~~ •..,...,._._,.., ....... A 17'aplicit
-ma IMlbod.thesianaJ111nc1uc1oc11n tbt ,_.. a.t • •
s ..
 -1\1(
ur in either the forward direction ~~,.
60 ~
9.
~~.
• • sirma
I 1

dif\'(tion.
· can occ
11ng

. bit is set in a packet moving in

or tht b
Oct
'
1 NETWORK LAVER
9.61

rd signahng, 8 . 1 t1le d'

In back"·• . . n This bit warns t ie source about th con'teer'llii • A leaky bucket algorithm shapes bursty traffic into fixed rate traffic by
• . the congc,t10 . e
opp0s1te to to slow do\\11. &esti averaging the data rate.
. s the source . o4
and 1nfonn . 8 bit is set in a packet moving in the d' I • Imagine a bucket with a small h~le at the bottom as shown in fig. 9.30.
rd signahog, • • 1rccr
In for,U. 1111s . . 1
s the destination about t ie congestion . The r 100 or
bit warn .
congesuon. !ides such as slowmg down the acknowled eceive,
· this case uses po gen1eni

~,, "'"""'"
Ill • S~
remore the congestion. • . d . d•fli
. trol policies use in 1 erent la" .., er,
C gest1on con
I use different policies or techniques to
9.9.3 on od 111 (variable rale)
I . rs of OSI m e . control
Different a) e Th ious policies used by data lmk layer, network I Ilic
~ongestion in subnet. e. var e table below : ayer~
transport layer are sho\\n mth Interface
containing
leaky Bucket
Policies bucket
Layer holds pad<elS

Data Link Layer I. Retransmission policy

Fixed flow
2. Out of order policy (Constant rale)
3. Acknowledgement policy

.
Network

4. flow control policy

Network layer J. Virtual circuit Vs Datagrams inside the subnet ' Fig. 9.30 (a) A leaky bucket with water (b) A leaky bucket with packets
2. Packet queuing and service policy
• The rate at which the water is poured into the bucket is not fixed and can vary
3. Packet discard policy
but it leaks from the bucket at a constant rate. Thus (as long as water is
4. Routing algorithm
present in bucket), the rate at which the water leaks does not depc:nd on th.:
S. Packet lifetime management .
rate at which the water is input to the bucket.
Transport Layer I. Transmission policy • Also, when the bucket is full, any additional water that CllfaS into the bucket
2. Out-of-«der cachins polioy spills over the sides and is lost.
3. Ackr.owJcdgemca poJicy
4. Flow CODllol pollc:y
• The _. conceptCIII bo app6ed to ,...
in .. llltWDlk (Sec fil. 9.30 (b))
Consider that data Is oomms tolo .. - . II Wlilllle .... Suppoae that a
S. TiDIHUl dcfamimtion IOlll'Ce sends data at 12 Mbpttlr.t . . . . ..,... . . ·•-tbr3 IIICClllds. The
9.9.4 Coqeation control aJpdtJuu tourceapln crmlta . . tta•ofttt5s farl'1 t - - - - - - ~of
1. Leaky Bucket AJcorWua 91CCOnds.68Mb dMabu•ll t frN.
• Itisatllffic ......._ ---..:_
......,.... - ~ that controls the amount and the rate of lbl :
f ~a leaky bucket ...-1a.r.1bt-.8owwll•I ...,.,..... TIius
constant flow 11 ""i 1 hId (Ill a,. U I)
 NETWORK LA YER 9.65

cotllltcr is 600. which is less th an the size of the next packet. The next three packets
camwt he sent. Thl'Y need to wnil for the next tick of the clock.

• packets cannot be
s~nt; must wait for the next tick
Countor
ooo
Counter
3400
Counter
6800
Counter
10,000

3000 2900 3000 2800 3400 3200

bytes bytes bytes bytes hytes bytes

Fig. 9.34. Leaky bucket implementation for variable_sized packets

2. Token bucket Algorithm

• The leak-y bucket algorithm allows only an average (constant) rate of data
flow. Its major problem is that it cannot deal with bursty data.
• A leaky bucket algorithm does not c~nsider the idle time of the host. For
ex■nple, if the host was idle for 10 second~ and now it is willing to sent data
It a vuy high speed for another IO seconds, the total ·data transmission will
lie clmded into 20 seconds and average data rate will be maintained. The host
illlavioan oadvanf- ae~idle for 10 seconds.
•
ff

~
-----~
le.• .... tllis ~ , _ tokonJ,ucket algorithm is used. A token bucket

••••• ""••
.
• cation of leaky bucket in which leaky

• In this algori1bm, a1t. ated at every clock tick. For a packet to

be transmitted, system must tbn(s) from the bucket. (See fig. 9.35)
• Thus, a token bucket algorithnt- idle hosts to accumulate credit for the
future in form of toke~.
• For example, if a system generate, 180 wk• in one clock tick and the host
• is idle for 100 ticks. The bucket wiD 10,000 tokens.
Now, if the host wants to send '°nsume all I0,000 tokens
at once for sending 10,000 cells
 ·•-s 1,,n11t, ·ti,,1M1ricn11-J ~rvk-,·"• 111i•11 tl .,, ~;:"~·~" 11•......- JO 13
·t ia,, tr I"''' l'"" , ~ 1(1~ .... .
• !:'Ill<' 11<'<'''~ . ., ~, ~t.1l>li$hins. o 111•w ,,,nn,,•ti,, 11 , In Nsf'l)RI ~YER
. i •· :\DII$ 11..0.11( \ I • su,•h or '""
.- ,,r rx 111111,1c. If 5000 byte, of dntn 1~ to he tmnsfcm:c.l ond the fi rst hytc is
• •
iirt\li , ii\ .. (Oti" t$a...l.1'\I I\' tdl whl\'h 1 l'l)l ls it h·1, • Si111aiL 1 1
t.y r<'tll•~ ,nr NC l\'rr.tr\$tlllltc.\ l.
lf1l\.~ -.1 " . • ~ n,11 ·"
l\\~ii1w 1111111 1,crcd
to.Oil I There nrc 5 segment~ of 10()() b) tc, cm:h then the sequence
11 L f d
ni..1:~~ • 'Ml ,r,1..JI in th<' ('l\\.'1~ ,f ,t11.1 u~tnsfi.-r. If s, .
" nu 111 ticr nf first ~q'.mrn t wnulJ be I0.001 . ~q11c1Kc nurn,i<:r o !l«On
1 II
• l\c11t!,(-h-'-~ ' ' . •l l\ 1•r~~ ~qii1tcn1w11ulJhell.00 l nncJ 50on.
. N1.'\t-k~-11 l1'l'l' "' ;111,,tlk-c lwt~ :m1l\'lllh:1t1S, that it haJ . s~ . ,
!l.~~ s . ..... . ;._ .... l\l int:'llmit,1f tht- st:1111.~,,f:111 ,, 1x,nc,,n 1,..., _JUst cta.u ' · •"· It provides stream data tran• fer
,:11! ~~ J.., ' 1~"'° ~'-111\ll, "~ A
• \\'ith stl\'tull dlllll tn111skr. TCP delivers an un,aruclurtd ~trcam of bytes

~.~.~~~-~ ............. .. .,
i.ttntitkd hy sequence nmnhcrs.
10.s. TUl{SPO RT............
...................... ..... This tirdlit)' is bcncliciol fo r the applications (processes) lx."Cause the
:-:::,:\!
-'· - ·
cr.!.'t. t.'ie ttr~'f'-'111 IJycr pn.11i..k'$
,~:,.~ ~ ,\.Yl."l(\.1 i._~~
t \\\) t)JX'S ,,f s,•ni1:cs :··~:···•····,
. c,,~ .
~...
•
P").:<'S$eS 11<"-'J not to chop d.11u into blocks before pa.ssing it to TC'P. Instead
TCP groups bytes into segments (a unit of d;ita at trJnsport la) cr is callc-d
s,·gm,·nt) nnd pass them to IP for ddi \'cty.
f..T t.:s. tr.ii\'111 ~tt J..·t:~ ti\\) Jit"tt-n:-nt 1'fl'tt~'1)ls Trm1sniission
!>-..~,, \''. t: f P' nl r~ Ds:::\f::i:n Pfl'll'-"-~ tU)P) CoQ~
3• If offers reliabilit
y
• TCP offers reliability by providing conn<-ction oric:mc:d, end to end rcliJbk
10.S.l Tn.nsmission Control Protocol
Jdivcl)' of P3rkets.
• lt is o~h: to ud1ien: this by S<.'\1u.:n.:ing bytes v.ith a forn1Ul:I
ud:m",·ledi;,·mcnt numhc:r tlmt indicates to destination the next b}1e of data
th<.' s,,ur,:c ,:x~-cts to n:cci,·e.
• B)1CS thnt are not neknowkdged \\ithin a sped tit.-d ti me ~ riod
are
rctr.msmitted.
• The reliability mcchanism of TCP allows J,:\'k,:s IL' Jcal 1\i th lost, dd a~'--d,
duplicate or misread p:ickets.
• A time-out mechanism allows de\iccs to detect lost plcl;.ets IDd rcqutSt
~.
4. It pro.we. etllcleat flow coatral
• TCP uses slidq window ~ 1D illfl1• r 11ow COIIIIQI.
• W'121l diis tbl 19CC1M1111aa Jll'OOIII . . . . . . . . . . . . . . . aumbcr
• kcareceiw...._4Ml.._lla..._..._
............. ...............
• '1111 . . f.lfCGNI . . . . . . . . . . . . . . . . tllt#t
• •-flll1 0a-a11111•" 6_,_.. _..._ ......
......., ...... ...Is:. . .
• TCP lllows .._his II I• lf#ill .... WWW
over a ainck ~
i I ••--Ir oit
 -.-vll.'t
10.16 . . ·cd indicate that TCP segment c ~'rtq
n set rt JS us 10 ontni11
LIRG, \\ en ' Jlllinter field is valid. s llrg
~er data and the urgen 1
oRf LAYER 10.17
~~, rRANSP
w . . set 10 I it means that the acknowledge
'I\ If ACK btl JS ' lllcnt 1 The various features of UDP are :
Al · . d ntains the sequence number of next byte lti111bc
lidd is ,·al1d an co exPccted r provides connectionless transport service to the applications.
I. 11
the~~ . ~ 2.
l)eing connectionless, its service is unreliable.
. b. . used to indicate to the receiver that the appli . • ~ J. It does no! offer flow control. It meartS receiver may overflow with
p1;ff This 11 1s . ca1r 011 t
• • d. nd has requested that tl11s segment of data entih. messages.
at the sen ing e rnay n .,
the receiving end and handed to the user entity 01 be
4. There is no error control mechanism in UDP except for the checksum.
_.J
bufferni aI as soon
possible. as This means that the sender does not known if message has been lost or
This bit is used for resetting the cormection. When this b.. duplicated.
RST. . d. I it is se1
I. the connection has to be cleared 1mme rnte y. to s. UDP is less complex and simple than TCP and easy to implement.
S\"N This bit is used for synchror,jzing the sequence numbers d . Because of its simplicity, UDP header contains fewer bytes and consumes
,. . uring th Jess network overhead than TCP.
connect10n. e
6. Being connectionless user datagrams (A unit of data handled by UDP) are
FIN This bit is set to I for closing a TCP connection in one dire 1.
· c ton. sent independently. There is no relationship between different user
8. Window size. This 2 byte field specifies the size of sender' . datagrams even if they are coming from same source process and going to
• ·1 bl c . . s receive
window (that IS the buffer space ava1 a e 1or mcommg data). A . the same destination process.
field is of 16 bits, maximum window size is 65,535 bytes. s this 7. User datagrams are not numbered.
9. TCP checksum. This 16 bit field contains the checksum. 8. UDP is useful in situations where reliability mechanism of TCP are not
IO. Urgent pointer. This 16 bit field is valid only if urgent pointer is set f necessary. The various situations where UDP can be used are:
The \'alue contained in this field is added to the value containe~o ~ , (i) UDP is used for transmission of digital voice over IP.
seq~nce number field to obtain the number of last urgent byte in the da · (ii) UDP is suitable for a process that requires simple request response
sectJon of the segment. . ta communication with little concern for flow control & error control.
IL Options. There can be 40 bytes of optional information in the TCP (iii) UDP is suitable transport protocol far muhicasling.
header. • (iv) UDP is used for ........... pn,ces.,es such as SNMP.
12. Data. This field contains the upper layer information. (v) UDP is med for .... mwe ....., paelOCOls u:h • Routing
10.5.2 User Datagram Protocol Infonnatiooh>tGcol (RIP).
• User D,.,,..,,.,...,
.._ ...., protocol {UDP) is a ~
protocol.
■-1:..1.,_
~
(vi) UDP can bo UNCl wilb DNS ia GIW to alt
lll)IOifio,IP......
Ille..._ - for a

• Like TCP, UDP also provide, Jllucea to . 10.s.a. 1 UDP

TCP, it docs not provide f l o w ~ commwucadon, A datapn fl 1he
• · lea,,__
It.15 connect100 trnlrQlldroJ medlani•-•
. ......,,
•
establishing a ~
UDPi,..,__,1.....,,
it..,._ die JJl'OCelSes to transfer data wf
EachUDPclataaraffl
t A UDP header Is
-.;n_,m.RFC1'6t 'I 10.1.a.a,_.,.,.-,
The UDP datqrlm format oontaina four ft.- (S. le 10.W}