International Journal of Computer Techniques - Volume1 Issue 2 Dec - 2014 ISSN :2394 -2231 http://ijctjournal.orgPage 1 Comparing the Performance of Real-Time Applications based on IPv4 and IPv6 NetworksVahid Ghoreishi1, Shahram Mohammadi2MSc graduate1, Assistant professor2 Department of electrical and electronic en gineering University of Zanjan Zanjan, Iran Abstract—With the expansion of Internet, the current protocol address space (IPv4) is faced with a shortage and the new generation Internet (IPv6) is raised in order to solve this problem. Also with increasing interest of users in real-time and multimedia applications, the Internet and how to implement these applications to improve the quality of service (QoS) has become a significant challenge. In this paper, a network with both IPv4 and IPv6 protocols is simulated by using OPNET Modeler and implementation of QoS is tested on it. Packet loss, end-to-end delay and jitter were evaluated as network performance metrics. The results showed that satisfying the needs of multimedia applications on IPv6 are slightly weaker than IPv4, but the benefits have evolved more and more and the difference is negligible. Also according to the methods used in terms of this study, the combination of DiffServ and MPLS QoS implementation is the best option for both Internet Protocol versions. Keywords—IPv4, IPv6, real-time applications, Quality of Service (QoS), BE, DiffServ, MPLSI.INTRODUCTIONNowadays, Internet is the most important and popular network of the world. But with the unexpected increase of number of users, the current Internet (IPv4) is not able to expand and due 32-bit address space, it is not possible to connect more than four billion hosts on the network. In 1991, IETF developed a new generation of Internet Protocol with a 128-bit address space, currently known as IPv6 [1] [2]. On the other hand, real-time and multimedia applications such as voice and video, from the perspective of an Internet user, have a high degree of importance. Thus improving the quality of service (QoS) and needs of the new generation of real-time applications should be considered. Quality of service of an application can be defined as the level of user satisfaction with using the program. Networks where QoS is not activated there, are known as Best Effort (BE). Differentiated Service (DiffServ) [3] and Multi- Protocol Label Switching (MPLS) [4] structures have been proposed by the IETF are useful f or guaranteeing QoS [5]. DiffServ architecture has features such as traffic classification and scalability and can easily handle a large volume of data [6]. In this mechanism, a bit is added to the packet header. By this bit, a packet that arrive a router, declare belonging to a particular class of traffic. Routers in the face of this bit behave a certain way. This manner known as Per-Hop Behavior (PHB) and during it the router needs more than one bit to make decisions [1]. IETF assigned six valuable bits of TOS byte in the IPv4 header and TC byte in the IPv6 header to the DiffServ Code Point (DSCP). DSCP is used for selecting a PHB by a packet [7]. Three types of PHBs have been introduced by IETF. Default PHB tries to act as BE and has no guarantee in typical IP networks [8]. Expedited Forwarding (EF) has minimum loss and delay, and can be accessed by giving high priority to packets of EF or implementing WFQ with giving more weight to these packets [1]. The third PHB is Assured Forwarding (AF) that is rooted in RED method. AF is composed from four classes and each class has three different probability values to packets drop [1] [9]. MPLS improves scalability and flexibility of IP network and by optimizing bandwidth minimize the congestion effect [10]. The main idea behind MPLS networks is to provide facilities for Forwarding Equivalence Classes (FEC). All packets of a FEC accept similar label and transmit only through a path called LSP. Routers allocate a FEC to each of packets before sending them and then map each FEC to next step. So next routers don’t need to analyze the packet header. It increases the speed of the network [4]. In addition, the resources needed along the way, are already stored. Shim header is added to the start of IP header to insert MPLS label. This header is 32 bit and composed of four parts. An important part is EXP (or QoS) that defines the class of service [1]. In this study a network is designed under two separate IPv4 and IPv6 protocols and QoS performance metrics including delay, jitter, packet loss rate and throughput of the links in the BE, DiffServ and combining of MPLS and DiffServ are tested. At the end the results are reviewed with a comparative style. The second part will point to related works. In the third part we will propose a method with its simulation. Section IV presents the results of the simulation, and finally a conclusion in Section V will b e presented. II.RELATED WORKSAfter IPv6 is proposed by IETF, attempts to compare this protocol with IPv4 were done and still continue. It not only helps to standardize IPv6, but leads to the evolution of the current version of Internet. [11] provided two same workstations under Windows 2000 and Solaris 8 and IPv6 and IPv4 on these operating systems are implemented. Based on testing, IPv6 has lower throughput and higher
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
International Journal of Computer Techniques - Volume1 Issue 2 Dec - 2014
ISSN :2394 -2231 http://ijctjournal.org Page 1
Comparing the Performance of Real-Time
Applications based on IPv4 and IPv6 Networks
Vahid Ghoreishi1, Shahram Mohammadi2
MSc graduate1, Assistant professor2
Department of electrical and electronic engineeringUniversity of Zanjan
Zanjan, Iran
Abstract—With the expansion of Internet, the current protocol
address space (IPv4) is faced with a shortage and the new
generation Internet (IPv6) is raised in order to solve this problem.
Also with increasing interest of users in real-time and multimedia
applications, the Internet and how to implement these applicationsto improve the quality of service (QoS) has become a significant
challenge. In this paper, a network with both IPv4 and IPv6protocols is simulated by using OPNET Modeler and
implementation of QoS is tested on it. Packet loss, end-to-end
delay and jitter were evaluated as network performance metrics.
The results showed that satisfying the needs of multimedia
applications on IPv6 are slightly weaker than IPv4, but the benefitshave evolved more and more and the difference is negligible. Also
according to the methods used in terms of this study, the
combination of DiffServ and MPLS QoS implementation is thebest option for both Internet Protocol versions.
Keywords—IPv4, IPv6, real-time applications, Quality of
Service (QoS), BE, DiffServ, MPLS
I.
INTRODUCTION
Nowadays, Internet is the most important and popularnetwork of the world. But with the unexpected increase ofnumber of users, the current Internet (IPv4) is not able toexpand and due 32-bit address space, it is not possible toconnect more than four billion hosts on the network. In1991, IETF developed a new generation of Internet Protocolwith a 128-bit address space, currently known as IPv6 [1][2]. On the other hand, real-time and multimediaapplications such as voice and video, from the perspectiveof an Internet user, have a high degree of importance. Thusimproving the quality of service (QoS) and needs of the newgeneration of real-time applications should be considered.Quality of service of an application can be defined as thelevel of user satisfaction with using the program. Networkswhere QoS is not activated there, are known as Best Effort(BE). Differentiated Service (DiffServ) [3] and Multi-Protocol Label Switching (MPLS) [4] structures have beenproposed by the IETF are useful for guaranteeing QoS [5].
DiffServ architecture has features such as trafficclassification and scalability and can easily handle a large
volume of data [6]. In this mechanism, a bit is added to thepacket header. By this bit, a packet that arrive a router,declare belonging to a particular class of traffic. Routers inthe face of this bit behave a certain way. This mannerknown as Per-Hop Behavior (PHB) and during it the routerneeds more than one bit to make decisions [1]. IETF
assigned six valuable bits of TOS byte in the IPv4 headerand TC byte in the IPv6 header to the DiffServ Code Point(DSCP). DSCP is used for selecting a PHB by a packet [7].Three types of PHBs have been introduced by IETF. DefaultPHB tries to act as BE and has no guarantee in typical IPnetworks [8]. Expedited Forwarding (EF) has minimum lossand delay, and can be accessed by giving high priority topackets of EF or implementing WFQ with giving moreweight to these packets [1]. The third PHB is AssuredForwarding (AF) that is rooted in RED method. AF iscomposed from four classes and each class has three
different probability values to packets drop [1] [9].
MPLS improves scalability and flexibility of IP networkand by optimizing bandwidth minimize the congestioneffect [10]. The main idea behind MPLS networks is toprovide facilities for Forwarding Equivalence Classes(FEC). All packets of a FEC accept similar label andtransmit only through a path called LSP. Routers allocate aFEC to each of packets before sending them and then mapeach FEC to next step. So next routers don’t need to analyzethe packet header. It increases the speed of the network [4].In addition, the resources needed along the way, are alreadystored. Shim header is added to the start of IP header toinsert MPLS label. This header is 32 bit and composed offour parts. An important part is EXP (or QoS) that defines
the class of service [1].In this study a network is designed under two separate
IPv4 and IPv6 protocols and QoS performance metricsincluding delay, jitter, packet loss rate and throughput of thelinks in the BE, DiffServ and combining of MPLS andDiffServ are tested. At the end the results are reviewed witha comparative style.
The second part will point to related works. In the thirdpart we will propose a method with its simulation. SectionIV presents the results of the simulation, and finally aconclusion in Section V will be presented.
II.
RELATED WORKS
After IPv6 is proposed by IETF, attempts to compare
this protocol with IPv4 were done and still continue. It notonly helps to standardize IPv6, but leads to the evolution ofthe current version of Internet. [11] provided two sameworkstations under Windows 2000 and Solaris 8 and IPv6and IPv4 on these operating systems are implemented.Based on testing, IPv6 has lower throughput and higher
delay than IPv4. In [12], delay and packet lossscale IPv6 are studied. Results showed thatIPv6 is higher than IPv4 and more packets areVOIP performance was tested by a softphonetworks based on IPv4 and IPv6. The values othe experiments show that there is little differethe performances of VOIP in the two versprotocol. Even the maximum jitter and packetIPv6 are slightly more than IPv4.
[14], [15], [16] and [17], implemented abased on DiffServ (EF) on a real IPv6 network6NET that has the lowest latency, jitter and pareal-time traffic. In [18] a DiffServ network caall path from source to destination and the IntSmapped to DiffServ by a mapping function. Afrom the results, the QoS requirements canclearly in this way. [19] studied DiffServ QoSunder the three service classes EF, BE andversions of IP in two separate test bedclassification with access list to use flow labelheader. The successful implementation of theproven that the flow label classification basedpossible within a crowd. In [20] DiffServ is usQoS, but since DiffServ cannot control traffic c
end-to-end paths, MPLS do this. The resultsPacket Delay Variation (PDV) in IPv6 is more t
III.
THE PROPOSED METHOD AND SIMULATI
A. The Scheme and Network Topology
Here's a simple computer network intendevideo conferencing applications as network trafinto account and can occupy the entire networBE, DiffServ and combining of MPLS aarchitectures were used at three separate moeffectiveness in providing end-to-end QoS weDesired network is investigated in two indepewith different traffics and architectures. At the felements are configured to support IPv4 and
case, to support IPv6 only. Fig.1 shows the netby OPNET Modeler 14.5 simulator.
Fig. 1. The network configuration in OPNET sim
B. Network Traffic
To test the method proposed in this study,multimedia traffic such as voice and video
Computer Techniques - Volume1 Issue 2 Dec
http://ijctjournal.org
rate in largethe delay inlost. In [13]e for LANtained from
nce betweenions of theloss rate of
QoS serviceconnected tocket loss forn be seen inerv traffic is
we can seebe achievedmechanismsBE on both
and usedfield in IPv6e tests haveon DSCP isd to supportongestion of
suggest thathan IPv4.
ON PROCESS
. Audio andic was takenbandwidth.
d DiffServe and their
e compared.ndent modesirst, networkin the other
ork created
ulator
two types ofonferencing
traffic is taken into account. For vwith PCM quality, has been chosfor transmitting a digital phone syPCM data rate is 64 kilobits per seof the frames is 10 ms, i.e., an audior 80 bytes and 100 frames willaudio used in EF_Source and EF_and to be close to the volume of tra13 audio applications are used sim
with low resolution is used for vpixel of the video, is equivaletransmission rate in bits per secoThe video application is locateworkstations (AF4x_Destination ato measure the impact of congesttraffic began to flow at a time anend of the simulation.
Backbone of the network wLER_In and LER_Out, is the mainMPLS. In this area, there are two quse of OSPF algorithm will divideequal between paths. At the end,8.2 Mbps on the links in this area is
C.
Design TestThe Test consists of three pa
influence of BE, DiffServ and Meach of these sections are impleIPv4 and IPv6. To prepare the teaddressed and then as OSPFvalgorithms are implanted for IPv4the backbone network. To apply Bof TOS and TC fields and consezero.
In the second part, the DSapplication goes to EF mode and iAF41, AF42 and AF42 values. Eused because of their appropriate
VOIP and video conferencing [22]AF42 and AF43, DSCP value b(100010), 36 (100100) and 38 (10PHBs allocate applications to sepavideo applications, each pair of woof the experiment those were the sAfter separation of traffic,transmitted to the backbone of ttheir priorities. This should be don(LER_In router). To prepare this(ACL) is defined on it, which helpand provides a classification baseOnce the traffic is identified baseand is classified and marked basinput interface of LER_In, it turns
mechanisms have been used for thWRED traffic classes are available.
The final section of test is the ion a test bed which DiffServ enabof the test, packets sent fromregardless of the routing algorithm
- 2014
Page 2
oice traffic, conversationn. PCM coding schemetem is discussed in [21].cond and the default sizeo frame’s size is 640 bitsbe sent per second. TheDestination workstationsffic on the other stations,ultaneously. Then, video
ideo conferencing. Eachnt to 9 bits and datad is equal to 1,382,400.
on three couples ofnd AF4x_Source). Hereion on the network, thed will continue until the
ich is located betweendomain of DiffServ and
uite similar paths and thethe traffic approximatelyxperimental data rate ofdefined.
ts, which in turn, studyLS on the network andented separately on thet bed, the interfaces are
and OSPFv3 routingand IPv6 respectively in
to the system, the valueuently DSCP should be
CP field in the audiovideo applications takesand AF4x services are
ess and acceptability in
. By applying EF, AF41,comes 46 (101110), 34110) respectively. These
rate classes, especially inrkstations in the first partame, will act differently.pplications should bee network according toe at the edge of the areaouter, first an access lists router to identify traffic
on the source address.d on the source addressd on DSCP field at thequeuing traffic. CBWFQ
is work. Next, apply the
mplementation of MPLSed on it. Under this parta predetermined path,
for each class of service depending on tapplication. To do this, two round trip aredown the MPLS domain to transfer voiceapplications. Then FECs are defined. In oaccording to the application, four types of FEthose are separated according to the DSCP in tIP packets. The second general definition fstructure is traffic body that characterize incoflow to LSPs and bound FECs. By definitio
traffic body profiles based on common PHBadapt MPLS to provide QoS. That must bebetween the EXP bits in the MPLS shim headfield of the IP header to make them understandother and classes of service are defined for MPModeler has a standard EXPPHB mapvalues in this study according to the guideligiven in Table 1. Finally, the general definitionbe called in two edge routers in the MPLS domshould be dedicated. In this experiment, the uused for the transmission of audio and firstconferencing application and two other videoare transmitted by lower path.
TABLE I. EXPPHB MAPPING
PHB EXP
EF 0
AF41 1
AF42 2
AF43 3
(c)
Fig. 2. End-to-end delay of audio packets in EF_Source a)
(c)
Fig. 3. End-to-end delay of video packets in AF41_Source
IV.
TEST RESULTS
The Results which we focus on their, are
Computer Techniques - Volume1 Issue 2 Dec
http://ijctjournal.org
he type ofade up andand video
ur network,s is defined
he header ofor a MPLSming trafficn FECs and
s, turn is toa mappingr and DSCPble for eachLS. OPNETing that its
nes [23] areabove mustin and LSPspper path ispair’s videoapplications
(a) (b)
BE, b) DiffServ, c) MPLS
(a) (b)
a) BE, b) DiffServ, c) MPLS
he so-called
QoS performance metrics. QoSdepends on the application thaapplications, may have different defollowing criteria are consideredother definitions can also be con[25].
●
Delay: shows the period bfrom the source to the dMostly expressed in terms o
●
Jitter: This measure hasmain definition introducedbetween the delays of two c
●
Loss: The percentage of dthe destination at specified t
D. Delay
In figures 2 to 5 audio anddelay at the source node can be seIPv4 transmit packets slightly fastis that in IPv6 networks, pareconstruction is performed ondestination nodes and packet paylconstant. But in IPv4 networks, senetwork nodes [26]. Another impor
of DiffServ to BE. The reason forspent in the DiffServ domain edgeand queuing of packets. This unethe DiffServ is not able to meetMPLS should be used to incrapplied in AF42 and AF43 serviceThis delay comes from overcrowdembedded to these two servicessignificant point is the stability of tto reduce the jitter.
(c)
Fig. 4. End-to-end delay of video packets i
(c)
- 2014
Page 3
parameters are defineduses it and different
finition for those. But thes the basis of QoS and
verted to this form [24]
etween data transferringstination until it arrive.f end-to-end delay.
several definitions. The jitter as the differencensecutive data units.
ata units that don’ reachime.
ideo packets end-to-endn. The results show thatr than IPv6. The reasonket segmentation and
ly at the source andoad in the entire path ismentation is done at the
tant point is further delay
this paradox is the timerouters for classification
xpected delay notes thatthe needs for QoS andased efficiency. MPLSalso increases the delay.
Fig. 5. End-to-end delay of video packets in AF43_Source
E. Jitter
To investigate the jitter on the network, soare used. In OPNET this measure in video apnamed as Packet Delay Variation (PDV). T jitters of experiments are observed in figures 6IPv6 is generally better than IPv4. This point
about the EF and AF41 services that are tranover our network. AF42 and AF43 services alsosituation before applying MPLS, but after theintended to be more crowded, IPv4 jitter is lesDiffServ create more jitter than BE. MPLS is aEF and AF41 PHBs improve jitter, but on tPHBs due to path congestion, this standard is in
(c)
Fig. 6. Jitter in EF_Source a) BE, b) DiffServ, c) MPLS
F. Loss
To check the amount of packet loss, receivbe addressed workstations and compare it withThe same as before, received traffic at the sourstudied to observe the packet loss rate. Resimulation are shown in Figures 10 to 13. Wiview we can see that in the absence of Trafficwith MPLS, IPv6 packet loss is less than
architecture allows separation of traffic pathsIPv4 loss. Here BE acts better than DiffServloss. Finally, as expected, EF and AF41 PHBlosses than AF42 and AF43. This returns to tprioritization.
(c)
Fig. 7. Jitter in AF41_Source a) BE, b) DiffServ, c) MPLS
smitted wellhave such aath they ares than IPv6.pplied in thee other tworeased.
(a) (b)
d traffic cansent traffic.e stations is
sults of theh an overallEngineeringIPv4. This
and reduceand has less
have fewereir inherent
(a) (b)
(c)
Fig. 8.
(c)
Fig. 9. Jit ter in AF43_Source a) BE, b) Di
(c)
Fig. 10. Traffic received in EF_Source a)
V.
CONCLUSI
The results show that the fourhave direct relationship to each othof one, another improve. Figuresthat IPv4 provides QoS requiremIPv6. But it is not too much diffeother benefits of IPv6 could makeAnother important result of theDiffServ architecture alone cannotime applications and it is not mucmight work worse. For this purmust be used in combinat ion to Dsatisfied as possible.
- 2014
Page 4
(b)
. Jitter in AF42_Source a) BE, b) Dif
(b)
ffServ, c) MPLS
(b)
E, b) DiffServ, c) MPLS
N
riteria of interest almoster and with improvementan be generally explain
ents slightly better thanrence between them andthe difference tolerated.experiment is that the