Advanced Network Technologies

8/18/2019 Advanced Network Technologies

1/52

School of Engineering and Design

Advanced NetworkTechnologies


2/52

Advanced Network Technologies

Summary:

Networking is one of the most fundamental areas of research. Advanced Network Technologies are being developed so that communication infrastructure can be made as

simple and effective as possible. This course consists of the study of communication modelsalong with its topologies and protocols. In this report we are going to focus on DifferentialServices !DiffServ" and #ultiprotocol $abel Switching !#%$S" along with the &uality of Services re'uired in the research part and then continue with the simulations using Network Simulator !ns(".

)e have carried out simulations which give us detailed information about the T*% and +D%traffics supported by the performance parameters such as throughput packet loss delay and

,itter. )e analy-e and plot results using the N+ plots. )e also look into the 'ueuemanagement re'uired for networking which would consist of different types of 'ueues such

as Droptail S/& and 0ED. )e finally round up by having an overview of DifferentialServices congestion windows and wireless node management involving both static anddynamic nodes.

(+niversity of Susse1 *andidate Number2 34567


3/52


Acknowledgement and Statement of Originality:

I hear by certify that the report on Advanced Network Technologies includes research part onDifferential Services and #ultiprotocol $abel Switching and conclusions from laboratoryobservations only. I would specially like to thank Dr. Elias Stipidis for providing me anopportunity to carry out my research under his guidance. I would also like to thank thelaboratory assistances who provided us with help at any time of the hour whenever re'uested. #ost of the work in this report is original and work of others has beenacknowledged with their appropriate name and links at the end of the report. I havesuccessfully gained a lot of valuable knowledge and e1perience which would help me infuture implementations in the same field as a part of personal interest developed whileworking on this pro,ect.

I would also like to take this opportunity to acknowledge

• Dr. Elias Stipidis.• #r. Aditya Deshpande.• #r. David Abdulmasih

*andidate Number2 34567

778678(676

#Sc in #odern Digital *ommunication Systems.

+niversity of Susse1.

9+niversity of Susse1 *andidate Number2 34567


4/52


Table of Contents:

7 Introduction.........................................................................................................................5( Theoretical :ackground......................................................................................................;

(.7 Differential Services !DiffServ"..................................................................................;

(.7.7 Differential Services Architecture........................................................................;

(.7.( *lassification by %er8


5/52


9.; Advanced Network Analysis.....................................................................................95

9.;.7 Throughput and *ongestion )indow of T*%8SA* and T*%80ENB............95

9.;.( )ireless Network Simulation............................................................................9=

3 *onclusion........................................................................................................................3(> 0eferences2........................................................................................................................39

4 Appendi1 : C Bbservation for Differential Services........................................................33

>+niversity of Susse1 *andidate Number2 34567


6/52


List of Figures:

/igure 72 #odel of DiffServ registration8domain8based wireless network architecture...........;/igure (2 DiffServ /ield.............................................................................................................=

/igure 92 )orking of #%$S system.........................................................................................7(/igure 32 NS8( animation showing different packet si-es.......................................................75/igure >2 T*% packets and +D% packets. There is packet loss due to bottleneck condition.. .7=/igure 42 T*%8+D% combined Traffics are shown with increasing number of Nodes............((/igure 52 Throughput for T*% +D% and T*% C +D% combined.............................................(9/igure ;2 Delay graphs for T*% +D% and T*% C +D% combined..........................................(4/igure =2 @itter for T*% +D% and T*%8+D% Traffic...............................................................(;/igure 762 Throughput and Delay comparison of 0ED Droptail and S/&.............................99/igure 772 #ultiple Transmissions of 9 wireless nodes...........................................................37



7/52


1 Introduction

In the first part of the report we are going to focus on Differential Services #ultiprotocol$abel Switching and &uality of Service needed for the system to run smoothly in hostileenvironments. )e will be discussing e1actly what do we mean by Differential services and#ultiprotocol label switching.

In the second part of the report we are going to analy-e and discuss various practicalsimulations carried out in laboratories using NS8( simulator. )e simulate various tcl scriptse1plaining T*% and +D% traffic. )e will learn how to use the network animator to viewgraphically all the simulations taking place. )e discuss about the different performance

parameters such as throughput packet loss delay and ,itter for managing 'ueues such asDroptail S/& and 0ED. )e will simulate fi1ed and adaptive routing protocols in placewhich are used in practical systems such as Differentiated services congestion windowalgorithms and wireless nodes either static or moving. At the end of the report we would becomfortable working in practical situations and handling hardware designed for different

Network topologies.



8/52


2 Theoretical Background

2.1 Differential Services (DiffServ)Differential Services is a class8based mechanism which is used for traffic management. It is'uite different from others in a way that it uses the principle of Traffic *lassification whereeach packet is divided into different groups before it is transmitted rather than simplereservation of packets. Differentiated services solve the traffic issue in I% networks by simpleseparation of packets.

2.1.1 Differential Services Architecture

Figure 1: The conceptual model of iffSer! registration"domain"based wireless network architecture #$%

;+niversity of Susse1 *andidate Number2 34567


9/52


2.1.2 Classification b !er"#o$ Behaviors (!#B)Differentiated services were introduced to provide end8to8end &uality of Services

!&BS". The packets are controlled by router by using different settings for per8hop behavior !%


10/52


9 *lass 9

( *lass (

7 *lass 7

6 :est effort

Table 1: 'recedence Le!el with description of classes #&%

The Assured /orwarding %


11/52


DS*% 7( DS*% (6 DS*% (; DS*% 94


12/52


2.1.& Advantages• DiffServ scales very well since connection8oriented states are not stored in the nodes.• There is no further delay or stressing of the nodes since DiffServ does not use

signaling 3 .• :ehavior of the network is easy to measure as there is no comple1ity in the network.

• It removes addition burden on the core network of imposing comple1 routingtechni'ues 3 .

2.1.' Disadvantages• It is difficult with DiffServ to support end to end 'uality of services.• *lassification of the packets can be e1pensive 3 .• If packets are transported over multiple DiffServ domains there is no guarantee that

they receive the same treatment in all domains 3 .• Since classification is based on priority mechanism on congestion low priority

packets may get dropped ( .

7(+niversity of Susse1 *andidate Number2 34567


13/52


2.2 ulti$rotocol abel S*itching ( ! S)

#ultiprotocol $abel Switching was a technology that was introduced into the market whichwould have the advantages of AT# Networks and would provide a solution for disadvantagesof the same system. #%$S works between layer ( and layer 9 in the BSI model. #%$S

belongs to the classification group of %acket8Switched Networks. #%$S is a mechanism usedin higher performance telecommunication networks which directs and carries data from onenetwork node to the ne1t 4 .

2.2.1 +orking of ! S

Figure &: )orking of *'LS system+ #,%

Initially when the I% packet enters into the network identification like a label is attached toeach packet at the edge of the network. This identification describes the virtual privatenetwork it is received from. The packet has to pass through various switch points beforereaching the destination where its identification is removed and the packet gets transformedinto its original state. The main task of #%$S is to provide traffic engineering which resultsin reduced traffic on the network. It also minimi-es the cost of the network. #%$Sforwarding scheme is based on the label information based !$I:". /or forwarding incoming

packets #%$S checks for the identification to be used in its $I: to determine outgoinginterface which is based on forwarding e'uivalence class !/E*" > .



14/52


2.2.2 Advantages of ! S

• #%$S is simple as compared to other systems as there is no need of managing table

comple1ity.• #%$S has good 'uality of service !&BS".• #%$S enhances the performance of operational I% networks• #%$S offers traffic engineering for optimi-ing performances > .• If voice and data can be added to #%$S it would reduce cost of the network as

compared to AT# or frame relay network • #%$S supports implementation of Girtual private Network > .• #%$S integrates I% and AT# in the network.

2.2.% Disadvantages of ! S

• An additional layer is re'uired for #%$S.• The router has to be programmed separately for identifying #%$S labels.



15/52


2.% ,ualit of Service for ulti-edia A$$lications

&uality of Service is one of the most important factors taken into consideration by the service provider while initially designing the system. &uality of service is directly proportional tocustomer satisfaction.

&uality of Service H *ustomer Satisfaction )ith increasing technology the needs of the common man also increases. The multimediaservices like Internet Gideo *hatting have now become a necessity on wireless devices suchas mobiles and laptops. The increasing demand of such services leads to bandwidth scarcitywhich results in reduced 'uality of service or non8reliable services. In such situations #%$Salong with differentiated services are used to implement &BS in I% networks. They increasesystem capabilities and reduce traffic in the network > . The 'uality of service layers aregiven as follows2 ;

&BS $ayers &BS %arametersApplication /rame rate si-e and resolution 0esponse Time Throughput Security

%rice and *onvenience.System :uffer si-e %rocess %riority Scheduling %olicy *aching %olicy Time

&uantum. Network :andwidth Throughput :it Error 0ate End to End Delay Delay ,itter

%eak Duration.

Some of these parameters at different levels can be included as in general parameters defined

as accuracy precision and timeliness. Timeliness Accuracy %recision !TA%" can together form a good criterion for &BS. #ultimedia networks support real8time data bulk data andstatistically multiple1ed data which make the traffic management in the network hard. Thenecessary traffic management components to support &BS are2 ;

7" Admission control2 The admission control component takes into account resourcereservation re'uests and the available capacity to determine whether to accept anew re'uest with its &BS re'uirements.

(" Scheduling2 The scheduling component provides &BS by allocating resourcesdepending on the service re'uirements. This re'uires mapping the user8defined&BS re'uirement to resource allocations for providing the service.

9" 0esource management2 &BS can be provided using over8provisioning of anetwork which increases the cost incurred by the provider. Efficient resourcemanagement is a cost8effective solution for the provider and it ensures thatapplications will get the specified &BS during the course of its e1ecution.

3" *ongestion control2 *ongestion control is re'uired to avoid anything bad fromhappening inside a network domain. Some applications may not follow thestandard protocol description and try to steal resources thereby deteriorating the&BS of other applications. #echanisms are needed to recover from congestionand control flows accordingly.

>" %olicing?Shaping2 +sers might send traffic at a rate higher than the agreement.%olicing is necessary to monitor these situations and shaping makes the trafficsmooth and reduces its variations over time.

7>+niversity of Susse1 *andidate Number2 34567


16/52


17/52


% aborator Si-ulations

%.1 Introduction to 0et*ork Si-ulator 0esearch of any kind in networking needs simulations to generate perfect practical scenarios.

Network Simulator is used to simulate such environments which may consist of Nodes whichcan be mobile units in the real world 0outers and most importantly it can generate data flowof any magnitude. Thus we can use this software to study bottle links traffic analysis of various types such as *:0 and +D%. )e can simulate practical environment and measure thethroughput delay and ,itter of the system using network simulator. )e can analy-e different'ueue management schemes along with routing protocols and DiffServ using different codesas accepted by NS8(.

• E1planation of some variable parameters such as Start and End time Delay $ink *apacity %acket Si-e that will be used throughout this report.

Start and 2nd time: It reflects the time according to the code when the simulation starts andwhen it finishes.

elay: Delay is the time taken by the link to transfer one packet from source to destination.

Link Capacity: $ink *apacity corresponds to the ma1imum tolerable traffic on the link. It

should be large in order to avoid losses but not large enough to avoid wastage of link bandwidth.

'acket Si3e: %acket si-e indicates the amount of data that a single packet can hold duringdata transfer from one node to another.

/ueue Length: It defines the si-e of the 'ueue. If 'ueue length is greater than average the probability of packet loss is less and vice8versa. The default 'ueue length in NS8( for Droptail is 76.

TC'45 ': T*% and +D% are traffic agents which correspondingly use /T% and *:0. /T%needs acknowledgement before the ne1t packet is sent while *:0 is connectionless servicewhich does not need any acknowledgement.



18/52


%.2 Si-$le 0S si-ulation scri$t for t*o nodesThis code creates two nodes for e1change of data traffic. *:0 traffic is used for data transfer from node6 to node7.

Figure (: 6S"$ animation showing different packet si3es+

The Start time and Stop time is 6.> sec and 3.> sec respectively for all observations.

Link Capacity elay 'acket Si3e Obser!ations

7 #b 76 ms >66 No %acket loss.7 #b 6.> ms 7666 %acket loss at 7.(5>7.( #b 6.> ms 7666 No %acket loss.

6.> #b 6.> ms >666 &ueue starts at 6.>7 and packet loss at 6.>4> #b 6.> ms >666 &ueue starts at 7.6463

>6 #b 6.(> ms >666 No %acket loss.

7;+niversity of Susse1 *andidate Number2 34567


19/52


Conclusions:

7" )hen link capacity is reduced and all other parameters remaining the same packetloss is possible depending on the value of link capacity. If link capacity is e1tremely

small then there is more packet loss and if link capacity is very high then there is no packet loss. Thus link capacity is a ma,or factor.


20/52


%.% Traffic Anal sis *ith TC! D!

%.%.1 Transfer of h brid TC! and D! traffic over bottleneck link.

This code shows the data transfer between four nodes forming a bottleneck. Thedifferent types of traffic sources used are T*% and +D% and corresponding agents are used totransfer traffic over the bottleneck link n( and n9. T*% transfer packets only after acknowledgement is received for the previous packet whereas +D% does not re'uireacknowledgement for data transfer. *:0 traffic is used for periodic fi1ed length messagewhere Gariable :it 0ate is an e1ample of voice video or internet traffic. The minimum valuefor the link capacity of bottleneck link should be greater than the sum of the link capacities of individual links.

Figure 7: TC' packets and 5 ' packets+ There is packet loss due to bottleneck condition+

The blue packets are T*% packets and the red packets are +D% packets. )e can also observethe acknowledgement returning to the source of T*% packets.


21/52


Link capacity elay

'acketsi3e

C8. packetrate C8. start FT' start /ueue limit

7.5 #b (6 ms 7666 7#b 6.7 7 76> #b (6 ms 7666 7#b 6.7 7 76

7.5 #b > ms 7666 7#b 6.7 7 >67.5 #b (6ms 7666 9#b 6.7 7 767.5 #b (6ms >666 7#b 6.7 7 >6

Conclusions:

7" )hen T*% traffic is used window si-e should also be taken into consideration. Evenif there is no packet loss T*% nodes always wait for the acknowledgement beforesending the ne1t packet. So the amount of packets the node can send before it has towait for an acknowledgment has to be taken into consideration which is given bywindow si-e. )indow si-e would be studied in more details in further e1periments.

(" T*% and +D% are independent traffics. T*% stops sending data when it stopsreceiving acknowledgement or when there is no data to send. The data rate for +D% isconstant. &ueuing reduces as soon as one of the sources stop sending data as there isless traffic in the bottleneck link. The available bandwidth is sufficient to transfer theincoming traffic without generating a 'ueue.

9" )hen bottleneck link capacity is increased there is no packet loss. )hen link capacityis reduced there will be packet loss depending upon other parameters. )hen link capacity is increased the bandwidth of the link increases so there is no packet lossand vice8versa.

3" )hen 'ueue limit is increased packet loss reduces as more packets can be 'ueued.This approach can be used for avoiding packet loss but cannot be used for real timecommunication as time re'uired for the system to clear the 'ueue cannot be predicted.The priority given to packets in the 'ueue will be a ma,or factor.

>" If delay is reduces and *:0 packet rate is increased then there will be packet loss inthe system. There is increased packet loss because the time allocated for packettransmission is now less. Thus if more packets being transmitted in less time willresult in packet loss.

(7+niversity of Susse1 *andidate Number2 34567


22/52


%.%.2 ulti$le 3ando- TC! traffics in a bottleneck link.

In this e1ample /T% or *:0 start time and end time for traffic is random which makes thenetwork more realistic. It is difficult to predict traffic on the link at any given time. Since thedata has been designed to route through one bottleneck link packets can be droppedcorresponding to various parameters for that bottleneck link.

• This is a figure of T*% traffic on 9 different nodes. In this figure we observe that 6 C 7is the bottleneck link. )e can also notice the acknowledgement that is sent back to theT*% packet source.

• This is a figure of +D% traffic with increasing number of nodes. There are > +D%nodes on the transmitting and receiving end with 6 C 7 being the bottleneck link.

((+niversity of Susse1 *andidate Number2 34567


23/52


Figure 9: TC'"5 ' combined Traffics are shown with increasing number of 6odes+

Conclusions:

7" A long 'ueue is formed for T*% packets as every node waits for an acknowledgementfrom the receiver and this can result in all nodes transmitting at the same instant someof the time. %acket loss does not take place in T*% but the windowing effect has to beconsidered for ma1imum throughput.

(" As number of nodes increase the pressure handing capacity of the bottleneck link reduces which leads to loss of packets. The bandwidth of the bottleneck link is thesame which is the main reason for packet loss.

9" )hen bottleneck link capacity is increased then packet loss reduces because now the bandwidth available is sufficient for all the packets to get through from transmittingnode to receiving node.



24/52


%.& !erfor-ance Anal sis of the S ste-

%.&.1 Through$ut for TC!4 D!4 TC!" D!

Throughput is one of the most important performance factors that describe thecapacity of the link to carry data. $ink *apacity has a ma,or influence on throughput whereas$ink Delay is not a ma,or influence on throughput.

76666 C No loss Throughput for TC' >66 C 66 C


25/52


26/52




27/52


28/52


%.&.% DelaIt is an important parameter while measuring &BS for real time applications. )hen link capacity is decreased available bandwidth of the system decreases and therefore delayincreases considerably and when link capacities are increased then delay reducesconsiderably. )hen traffic is changed from +D% to T*% delay decreases and remainsconstant because +D% does not need acknowledgement whether the packet is receivedsuccessfully or not it ,ust keeps on sending packets at a constant rate.

T*% Delay +D% Delay

T*% C +D% Delay

Figure


29/52


Conclusions:

7" )hile we are using T*% agent we are using /T% applications. As the number of data packets goes on increasing and the link gets loaded delay goes on increasing until thesystem cannot accept any more packets and finally there is packet loss. )hen packetloss takes place throughput and delay of the system falls. )hen packets start buildinga 'ueue again then the delay builds up again gradually till packet loss takes place.

(" )hile we are using +D% agent we are using *:0 applications i.e. constant bit rate.Therefore delay remains constant after reaching saturation as the numbers of incoming packets remain constant.

9" )e can see a combination of increasing delay and constant delay for T*% C +D%traffic. Delay remains constant as soon as +D% start transmitting at 7.6 sec. The graphis a combination of individual graphs of T*% and +D%.

(=+niversity of Susse1 *andidate Number2 34567


30/52


%.&.& 5itter

@itter is the measure of the smoothness of the data transmission so it should be as low as possible. @itter is caused because of the data being divided into packets in a packet switchnetwork and each packet taking a different path from transmitter to receiver. If link capacityincreases bandwidth of the system increases congestion decreases and as a result @itter decreases. If number of nodes increases then congestion in the system increases because of which @itter also increases.

T*% +D%

Figure =: >itter for TC' 5 ' and TC'"5 ' Traffic+



31/52


Conclusions:

7" In T*% traffics acknowledgement is sent by the receiver which denotes that the packet was received successfully. If the packet taking different routes is lost then thetransmitter transmits it again. Thus e1plaining the @itter graph for T*% traffic whichshows almost typical value of @itter i.e. 6.6696 !from raph".

(" In +D% traffic no acknowledgement is sent by the receiver. Thus the transmitter bombards the input link with packets continuously which take different routes all thetime for reaching the destination. Thus different routes can lead to different time andorder of arrival at destination. Thus ,itter is a ma,or problem for connectionlessservices.

9" If the link capacity of the system is very high !the bandwidth of the system is veryhigh" the packets are transferred very 'uickly through the network and congestion isvery less. Thus packets can take the same route towards the destination which appliesthat the packets are received at the receiver in proper se'uence as transmitted by thetransmitter. Thus @itter reduces when link capacity of the system is very high.



32/52


%.' ,ueue anage-ent.

%.'.1 Dro$tail and Stochastic 6air ,ueuing (S6,)

)e have to compare two 'ueuing policies in this e1ercise. )e compare the throughputs andthe packet loss to distinguish them.

5 ' 'ackets:

$ink *apacity2 7#b %acket Si-e >66.

Droptail S/&

/igure 7(2 Throughput comparison of Droptail and S/& for +D% packets.

Comparisons:

.O'TA0L SF/Throughput decreases more as loss of packetsare more in Droptail.

Throughput goes on decreasing as number of packet loss goes on increasing.

%acket loss2 Sent 736( loss 369. %acket loss2 Sent 736( loss (7=.7st packet loss at 7.993 sec. 7 st packet loss at 7.733 sec.

9(+niversity of Susse1 *andidate Number2 34567


33/52


$ink *apacity2 6.> #b %acket Si-e >66

Droptail S/&

.O'TA0L SF/%acket loss2 Sent 736( loss 497. %acket loss2 Sent 736( loss (7=.7st packet loss at 7.6>= sec. 7 st packet loss at 7.733 sec.

*omparison for link *apacity2 6.> #b %acket Si-e 7666.

.O'TA0L SF/%acket loss2 Sent =59 loss 4=4. %acket loss2 Sent =59 loss 4(=.7st packet loss at 7.77 sec. 7 st packet loss at 7.(5; sec.

• Throughput goes on decreasing as packet loss goes on increasing. At all timesThroughput for S/& is higher than that for Droptail.



34/52


TC' ; 5 ' 'ackets Combined:

Droptail S/&

/igure 792 Throughput comparison of Droptail and S/& for T*% 8 +D% packets. !T*%"

Comparisons:

$ink *apacity2 7#b %acket Si-e >66.

.O'TA0L SF/Throughput decreases more as loss of packetsare more in Droptail.

Throughput goes on decreasing as no2 of packet loss goes on increasing.

%acket loss2 T*%2 Sent =3; loss (.

+D% 2 Sent =3; loss ;(

%acket loss2 T*%2 Sent 769; loss 5.

+D%2 Sent 769; loss (=9.7st packet loss at 7.>9= sec. 7 st packet loss at 7.(5= sec.

Conclusions:

7" )e observe that throughput for S/& is higher than Droptail at any instant. Droptail is based on first in first out policy while S/& follows a random flow of traffics in the'ueue and hence it provides better 'uality of service to the link as compared to

Droptail policy.

(" The 7 st packet loss is always in S/& but the total number of packets lost is more inDroptail. !from animator and awk script" In Droptail a 'ueue is formed after the link gets saturated. If more packets try to enter the 'ueue then the ma1imum 'ueue length

packets are dropped. In S/& packets are classified into flows and then assigned to a'ueue that is specifically dedicated to that flow. Thus every flow is given a fair chanceto transmit its data. If any of the 'ueue is e1ceeded packets are dropped.

9" To observe graph of throughput in T*% C +D% script flow ids are to be used toidentify T*% and +D% traffics.



35/52


%.'.2 3ando- 7arl Detection (37D)

0ED is an improved 'ueue management method for transmission over the congested

network. It helps to avoid global synchroni-ation. If buffer is almost empty all incomingtraffic is accepted. As the 'ueue grows probability for packet loss grows too.

Throughput Comparisons:

0ED Droptail S/&

elay ?raph Comparisons:

0ED Droptail S/&

Figure 1@: Throughput and elay comparison of .2 roptail and SF/+

Conclusions:

7" Delay of 0ED 'ueue is comparatively smaller and more variable.(" The throughput of 0ED is comparatively more stable and better as compared to S/&

and Droptail.9" In 0ED a random packet is dropped as soon as the probability of the 'ueue getting

saturated starts to go up.



36/52


%.8 3outing !rotocolsIn this e1ercise we analy-e the routing schemes for various networks. 0outing schemes isdivided into two main routing algorithms i.e. /i1ed and Adaptive 0outing. )e analy-e the

system by taking a look into the Throughput and Delay graphs.

Start of simulation /i1ed 8 after link break. Adaptive 8 after link break.

Throughput:

/i1ed 0outing Adaptive 0outing

.outing type packet transmitted 'acket loss

/i1ed routing >3; 7

Dynamic 0outing 766; 7



37/52


elay:

/i1ed 0outing Adaptive 0outing

/igure 7>2 Throughput and Delay for /i1ed and Adaptive 0outing.

Conclusions:

7" In fi1ed routing the packets are dropped and further packets are stopped fromtransmitting till the 'ueue is restored at (.6 seconds. If we are using T*% traffic no

acknowledgement would be returned back to the transmitter so no more packetswould be sent. If +D% traffic is used packets would be continuously sent and all

packets would be lost.

(" In adaptive routing the packets were routed through another link thus maintaining theoverall throughput of the system. The throughput fluctuates when there is damage tothe link and then rises again when all packets are re8routed whereas in fi1ed routingthroughput goes to -ero and then rises when the link is reconnected.

9" Thus both types of routings can be used according to the situation in the practicalenvironment. It depends on the &BS e1pected from the system. Adaptive routing isthe best option as it shows overall better performance in throughput then fi1edrouting.



38/52


%.9 Differential ServicesDiffServ is a techni'ue used to allow certain traffic more priority over others for better &BS.There are different types of policies which states how a particular connection should be

metered marked or policed during the simulation. The different policies are TS)(*#TS)9*# Token :ucket Single 0ate three colors marking and two rate three colorsmarking.

Obser!ation: Appendi1 8 :

Comparison:

Token 8ucket TS)$C*The throughput is little less than TS)(*# atevery stage.

The throughput is a little more at each andevery instant.

The number of packets policed from code point 76 to 77 and (6 to (7 is very high atevery stage i.e. (==;4.

The number of packets policed from code point 76 to 77 and (6 to (7 is very low atevery stage i.e. 333

There are some packets even though less thatthe packet gets transferred after being policedto 77 from 76 or (7 from (6.

No packets are transferred once it gets low priority i.e. 76 to 77 and (6 to (7.

The number of early drop packets is more atall times as com pared to T)S(*# i.e. 9776

The number of early drop packets is less atall times as compared to T)S(*# i.e. ==9

Conclusions:

7" )hen we change the parameters like cir cbs and rate to five times the original valuewe can observe that the total number of packets go on increasing but the total number of transfer packets remain the same.

(" The number of packets lost increases because the link capacity is kept the same whilethe other parameters are changed. The bandwidth and the capacity of the inputchannels have increased but the capacity of bottleneck link is still the same so there isincreased packet loss.

9" As we have increased the capacity of the number of packets that can be accepted thenumber of packets policed to a lower capacity is very low.

3" If we reduce the value of cir cbs and rate by five times the original value we canobserve that the total capacity is reduced but the link capacity is good enough totransfer the data packets. The link capacities of input channels have been decreasedand the bandwidth of the bottleneck link is sufficient to transfer all the packets to their respective destinations. Therefore there are no losses at all.

9;+niversity of Susse1 *andidate Number2 34567


39/52


9=+niversity of Susse1 *andidate Number2 34567


40/52


%.: Advanced 0et*ork Anal sis

%.:.1 Through$ut and Congestion +indo* of TC!"SAC; and TC!"370<

In this e1ercise we study about the different techni'ues for implementing T*% protocol. )ecompare the different techni'ues with the help of parameters such as window si-e delay

packet loss and Throughput.

TC' type )indowsi3e

elay Throughput Total numberof packet

transmitted

'acket loss

6ew .eno (3 3 ms =;4.3 kbps 7(49 74

.eno (3 3 ms ;;3.; kbps 773( 7>

Sack (3 3 ms ==4 kbps 7(5> 74

6ew .eno >6 3 ms =;6.; kbps 7(44 (>

.eno >6 3 ms ;;7.4 kbps 77>( (3

Sack >6 3 ms ==(.; kbps 7(;( (>

6ew .eno (3 76 ms =54.; kbps 7(>( 7(

.eno (3 76 ms ;5=.( kbps 779> 77

Sack (3 76 ms ==(.; kbps 7(5( 79

6ew .eno 76 3 ms ==4 kbps 7(;( 74

.eno 76 3ms ==4 kbps 77>9 79

Sack 76 3 ms ==4 kbps 7(=4 74

6ew .eno (3 7 ms =5(.4 kbps 7(54 (>

.eno (3 7 ms ;;; kbps 77>9 (3

Sack (3 7 ms ==;.3 kbps 7(;9 (>

Table &: Table for comparison of )indow si3e elay and Throughput+



41/52


42/52


43/52


44/52


45/52


& Conclusion

Thus we have completed a detailed study about Differential Services #ultiprotocol

$ayer Switching and &uality of Service needed for the system to run efficiently in practicalenvironment. Advanced technologies are still being developed to increase the 'uality of services simplicity and the receiver output. The concept of wireless communication i.e.mobile nodes is being taken to a different level all8together with the introduction of Ad8hocnetworks. )e have completed simulations on T*% and +D% traffic using NS8( simulator. )ehave detailed observations and conclusions about the different performance parametersrouting algorithms and 'ueues such as Droptail S/& and 0ED.



46/52


' 3eferences

7 0222 685;6985(648=?67? 75.66 J (667 C Admission *ontrol over Assured /orwarding%39?tk5>5?technologiesMtechMnote6=7;4a66;66=3=f(.shtml

3 Authors2 $ars :urgstahlerInstitut f r Nachrichtenvermittlung und Datenverarbeitung !IND" +niversitOtStuttgart and eorg *arle /B +S :erlin

Name of )ebsite2 www.ikr.uni8stuttgart.deDate last viewed2 96 th November (66=.+0$2

http2??www.ikr.uni8stuttgart.de?*ontent?&uasar?publications?#9.pdf

> 0222 674984;63?==? 76.66 J 7=== 8 #%$S Advantages for Traffic Engineering byP eorge Swallow *isco Systems.

4 *isco System *ooperation.inc $ast +pdated Date2 7> /eb (66;. Name of the website2 www.cisco.comDate last Giewed2 96 th November (66=.+0$2http2??www.cisco.com?en?+S?tech?tk394?tk3(;?tsdMtechnologyMsupportMprotocolMhome.html

5 #%$S EQ%E0TS

Name of the website2 www.mpls8e1perts.com

Date %ublished2 (66; R Date last viewed2 96 th November (66=.

+0$2 http2??www.mpls8e1perts.com?how8mpls8works.html

; 0aid $aboratories

Name of website2 www.purdue.com

Date last visited2 96 th November (66=.


http://www.cisco.com/http://www.cisco.com/en/US/tech/tk543/tk757/technologies_tech_note09186a00800949f2.shtmlhttp://www.cisco.com/en/US/tech/tk543/tk757/technologies_tech_note09186a00800949f2.shtmlhttp://www.ikr.uni-stuttgart.de/http://www.ikr.uni-stuttgart.de/Content/Quasar/publications/M3.pdfhttp://www.cisco.com/http://www.cisco.com/en/US/tech/tk436/tk428/tsd_technology_support_protocol_home.htmlhttp://www.mpls-experts.com/http://www.mpls-experts.com/how-mpls-works.htmlhttp://www.purdue.com/http://www.cisco.com/http://www.cisco.com/en/US/tech/tk543/tk757/technologies_tech_note09186a00800949f2.shtmlhttp://www.cisco.com/en/US/tech/tk543/tk757/technologies_tech_note09186a00800949f2.shtmlhttp://www.ikr.uni-stuttgart.de/http://www.ikr.uni-stuttgart.de/Content/Quasar/publications/M3.pdfhttp://www.cisco.com/http://www.cisco.com/en/US/tech/tk436/tk428/tsd_technology_support_protocol_home.htmlhttp://www.mpls-experts.com/http://www.mpls-experts.com/how-mpls-works.htmlhttp://www.purdue.com/


47/52


+0$2 http2??raidlab.cs.purdue.edu?papers?editor.pdf

= &uality of Service for I% Gideoconferencing Engineering )hite %aper SubhaDhesikan.

76 0222 " 7>9487(;3?6>? (6.66 J (66> C *ross $ayer Design of AD8hoc Networks for0eal Time Gideo Streaming by E0I* SETTBN TAESAN KBB QIAB&IN L


48/52


8 A$$endi= B >


49/52


(" TS)(*# 8 *I0 7666666R *:S 766666R rate 3666666

3=+niversity of Susse1 *andidate Number2 34567


50/52


>6+niversity of Susse1 *andidate Number2 34567


51/52


9" Token :ucket 8 *I0 >666666R *:S >66666R rate (6666666All parameters are increased to five times the original values for both the nodes.

>7+niversity of Susse1 *andidate Number2 34567


52/52


3" Token :ucket2 *I0 >66666R *:S (666R rate ;66666All parameters are reduced by five times the original values for both the nodes.

Advanced Network Technologies

Documents