Novel Mechanisms for Quality of Service Improvements in ... · Novel Mechanisms for Quality of Service Improvements in Wireless Ad Hoc Networks ... nidal}@mutah.edu.jo Abstract: -

Novel Mechanisms for Quality of Service Improvements in WirelessAd Hoc Networks

MOHAMMAD S. SARAIREH, NIDAL A. AL-DMOURComputer Engineering Department

Faculty of Engineering, Mutah University, POB 7Mutah 61710

JORDAN{m_srayreh, nidal}@mutah.edu.jo

Abstract: - This paper proposes two mechanisms for improving the IEEE 802.11 DCF protocol performanceand for enhancing QoS in IEEE 802.11 wireless LANs. These are Ratio based and Collision Rate Variation(CRV) schemes. The Ratio based scheme uses the collision rate value of the current and the past history of thenetwork conditions to adaptively adjust the Contention Window (CW) size for each individual station. TheCRV scheme also employs the collision rate and collision rate variation values to adjust the CW value locallyfor each individual station based on the current and previous network conditions. The aim of developing theseapproaches is to reduce the probability of collisions among the contending stations in a heavily loaded networkin an attempt to improve QoS in IEEE 802.11 DCF protocol. The proposed schemes are evaluated andcompared with the conventional IEEE 802.11 DCF and the Exponential Increase Exponential Decrease (EIED)schemes. The evaluation is carried out for different scenarios (i.e. different traffic load and different networksize) using the ns-2 simulator. The metrics used in the evaluation are delay, jitter, throughput, packet loss,protocol efficiency, and collision rate. The results indicate that the best performance is achieved by using theproposed schemes with superiority for the CRV scheme in most cases.

Keywords: - Quality of Service (QoS), IEEE 802.11 MAC protocol, Network Simulator, Contention Window(CW), Distributed Coordination Function (DCF), and Wireless Networks.

1. IntroductionWireless communications is a technology that isbecoming an important feature of many aspects ofour daily life. Not only are computer networksbecoming mobile, many devices will have one orseveral wireless interfaces such as laptops, cameras,and phones [1]. In some cases these devices andeven fixed stations wish to communicate with eachother without requiring an infrastructure. In thesecases, there is a need for ad-hoc wireless networksto provide an effective network communicationbetween different wireless devices. This type ofnetwork has a number of applications such asconferences, emergency operations, and militaryoperations [2] and [3].

Usually every device in ad-hoc wireless networkis able to communicate with every other devicewhen all devices are spread around a relativelysmall geographic range. The devices need to bewithin the transmission range of each other in orderto be able to establish a direct communication andto compete with each other to access the wirelessmedium. If the devices are out of the transmissionrange of each other due to lack of transmissionpower, long distance between the wireless devices,

interference, noise or due to mobility, it becomesessential to have an intermediary node betweenthem [3]. This node results in a multi-hop ad-hocnetwork. In this type of network a routing protocolis required and the Medium Access Control (MAC)protocol has to share the media with fairnessbetween different devices for different applications[4] and [3].

The function of the MAC protocol is to provideefficient and fair sharing of medium among allstations in the network. In wireless networks, MACprotocols can be categorized either as distributed orcentralized protocols [5]. Distributed wirelessnetworks, also called ad-hoc networks, are wirelessstations communicating with one another withoutany need for central administration. Centralizedwireless networks, are extensions to wired networksand have Access Points (AP) that act as theinterface between wireless and wired networks. TheAP polls the stations before assigning access rightsin turn and a station is only permitted to send whenit is allocated the right to do so. An example of acentralized protocol is the IEEE 802.11 PointCoordination Function (PCF) [6]. Distributedprotocols are contention algorithms that permit

WSEAS TRANSACTIONS on SYSTEMS Mohammad S. Saraireh, Nidal A. Al-Dmour

ISSN: 1109-2777 209 Issue 7, Volume 10, July 2011

stations in ad-hoc networks to be able tocommunicate according to the Carrier SenseMultiple Access with Collision Avoidance(CSMA/CA) mechanism. The IEEE 802.11Distributed Coordination Function (DCF) is oneexample of a distributed protocol. Other examplesinclude, HIPERLAN [35] and MACAW [36].

The main area of concern in this paper is topropose two mechanisms for improving the IEEE802.11 DCF protocol performance and forenhancing QoS in IEEE 802.11 DCF scheme [7].These are Ratio based and Collision Rate Variation(CRV) schemes.

In this paper, a number of related studies aredescribed in the next section. Section 3 introduces adetailed description of the Ratio based and CRVschemes. The simulation model is presented insection 4. The results obtained are analyzed anddiscussed in section 5. Conclusions are given in lastsection 6.

2. Related WorkSeveral algorithms that dynamically change thevalue of CW to improve the performance of theIEEE 802.11 DCF protocol have been proposed andare described in [7-18]. For instance, in [17], theLinear/Multiplicative Increase and Linear Decrease(LMILD) backoff algorithm is presented. In LMILDscheme, colliding stations increase their ContentionWindow multiplicatively, while other stationsoverhearing the collisions increase their contentionWindow linearly. After successful transmission, allstations decrease their CW linearly. An adaptiveDCF scheme was proposed in [18]. The proposedapproach is based on adjusting the backoffprocedure based on the knowledge of collision andthe freezes time the backoff timer of the stationexperiences. The study showed that, the proposedscheme outperformed the IEEE 802.11 DCF schemein term of throughput.

Several recent studies have improved theperformance of IEEE 802.11 DCF by eithermodifying the CW or adjusting the value of InterFrame Space (IFS). For instance the variation of theArbitrary Inter Frame Space (AIFS) betweenstations leads to a lower probability of collisionsand a faster progressing of backoff counter asreported in [19]. The author in [20] presented theAIFS as a technique for providing servicedifferentiation between different classes in the IEEE802.11e protocol. In [21], the length of DIFS wasadopted as a differentiation mechanism. In their

scheme, the DIFS length was calculated based onthe ratio of estimated transmission rate to the totaltransmission rate. Their scheme imposed majormodifications to the IEEE 802.11 DCF scheme inwhich the single queue was split into two queues.Their results showed that using variable length ofIFS, service differentiation can be achieved. In [22],the adjusted IFS parameter with other parameterssuch as quantum rate and deficit counter was usedto provide QoS mechanism. Their results showedthat using an adjusted IFS length, QoS could besupported. In [23], the authors proposed a methodfor optimizing MAC parameters in the EDCFprotocol, such as CW and DIFS. The proposedmethod improved throughput and delay ascompared with the IEEE 802.11e. However, thismethod was based on storing several networkconfigurations using database which imposed highcomputations.

Most of the discussed schemes require exchangeof information between stations. Moreover, theyrequire a sophisticated computation as the case in[24-26] and [23]. Other schemes impose majormodifications to the structure of the IEEE 802.11DCF as the case in [27-30]. Furthermore, moststudies only consider one or two of the QoSparameters. Moreover, they only depend on thecurrent conditions of the network withoutconsidering the past history. In this chapter, a Ratiobased and CRV schemes are proposed to overcomethese shortcomings aforementioned. They are assimple as the BEB to implement while significantlyoutperformed the IEEE 802.11 DCF and the EIEDschemes.

The term Quality of Service (QoS) has beenmentioned several times in this paper. This term iswidely used but with a variety of meanings andperspectives. For instance, RFC 2386 defines QoSas a set of service requirements to be met whiletransmitting data packets from the source to thedestination [37]. Another definition is that QoSrefers to the ability to provide a level of assuranceof data delivery and to provide a set of measurableservice attributes in terms of delay, jitter,throughput, and packet loss over the network. QoScan also be defined as the ability of networkcomponents, such as a host and application toprovide some consistent level of ensuring datadelivery over the network with different levels fordifferent classes of traffic [38]. In this paper, QoSrefers to the ability of the network of providing thedesired QoS requirements in terms of delay, jitter,throughput, and packet loss for the transmittedapplications.



3. Approach DescriptionThe basic IEEE 802.11 DCF protocol adjusts its CWvalue based on the current state of transmission, i.e.it doubles the CW value upon unsuccessfultransmission and resets to the CWmin uponsuccessful transmission [4]. The DCF scheme doesnot consider the past history of the network or thereadily available information. For the EIED scheme,EIED ( ri , rd ) is used to denote the amount ofincrease and decrease in the CW size aftersuccessful and unsuccessful transmission. Ifcollision occurred, the new CW is increased by themultiplication factor ( ri ) and after successfultransmission, the new CW is decreased by themultiplication factor ( rd ). In this chapter, the valueof 2 is chosen for each ( ri ) and ( rd ) as one of thepossible cases of the EIED scheme [31]. The EIEDscheme is used in the performance comparison,because it is not as aggressive as the basic IEEE802.11 DCF when the CW is reset to CWmin aftersuccessful transmission.

In order to make the protocol behave correctly,the CW value should be dynamically adjusted toadapt to the dynamic changes in the number ofcontending stations and in the amount of traffic overtime. This can be carried out by tuning the CWvalue after each successful and unsuccessfultransmission. This adjustment is carried out locallyfor each station at runtime. A detailed explanationof how the CW value is adjusted is given in thefollowing sections.

3.1 Radio Based Scheme (Case of SuccessfulTransmission)After each successful transmission, the DCFmechanism resets the CW of the station to its CWmin(i.e. CWnew = CWmin) ignoring the networkconditions. This action by the successful stationcauses frequent collisions especially when thenetwork is very large and heavily loaded because ofsmall value of CW. This agrees with the fact thatwhen a collision occurs, a new one is likely to takeplace in the near future since the collided packetrequires retransmission which causes extraoverhead. For this reason the Ratio based and theCRV schemes are proposed in order to mitigatebursty collisions. In the Ratio based scheme, the CWsize is adaptively adjusted based on computing thecurrent collision ratio for each station, sincecollision can provide a good indication about the

level of contention in the network. The currentcollision ratio is computed using the number ofcollisions and the number of successfullyacknowledged transmissions extracted from thehistory window ( wi ) as shown in Equation 1. Thehistory window ( wi ) is a sliding array that containsnumber of sent packets including part of the history.

])[(])[(])[(][

NsuccessfulNumNcollisionNumNcollisionsNumNR

wiwi

wiwicurrent

Where, ( ])[( NcollisionsNum wi ) is the numberof collisions for a station ( N ) that is extracted fromthe history window ( wi),( ])[( NsuccessfulNum wi ) represents the number ofpackets that has been successfully acknowledgedfor a station ( N ) that is extracted from the samehistory window ( wi ), ( ][NRwi

current ) is the current

collision ratio of a station ( N ). The ( ][NRwicurrent )

value is computed based on the number of collidedpackets and the number of successfully receivedpackets that are extracted from the history window( wi). The ( ][NRwi

current ) value is always in the rangeof [0, 1].

In order to maintain a continuous knowledgeabout the past history of the transmission the slidingwindow ( wi) is adopted. Furthermore, to reduce orto alleviate the random fluctuations in the computed( ][NRwi

current ) an Exponentially Weighted MovingAverage (EWMA) is used to smooth the series ofcollision ratios (i.e. ][NRwi

current value) as given inEquation 2 [32].

1)1( wiaverage

wicurrent

wiaverage RRR

Where the ( wicurrentR ) denotes the current or

instantaneous collision ratio for a station ( N ), ( )stands for a weighting factor which determines thememory size used in the average process, ( 1wi

averageR )represents the previous average collision ratio thatis computed form the previous history window( 1wi ), while ( wi

averageR ) is the average collisionratio at the current history window ( wi ).

The instantaneous collision ratio ( wicurrentR ) and

the average collision ratio ( wicurrentR ) are calculated

(1)

(2)(2)



based on the size of the total number of packets sentin ( wi ). The size of ( wi ) is selected not to be largein order to get a reasonable estimation about thenetwork status. However, it should not be too smallin order to get sufficient knowledge about thereadily available information of each individualstation. Moreover, using a sliding window ensuresthat the system always keeps a continuous trackingfor the history of the total number of packets sent.However, the size of ( wi ) and the weighting factor( ) are selected according to an extensive set ofsimulations carried out with several networktopologies and different traffic loads. In order toachieve a tradeoff value between throughput anddelay and in order to provide a good balancebetween removing short term fluctuations impactand capturing long term trends. Upon obtaining thevalue of ( wi

averageR ) described in Equation 2, the newCW size for a station ( N ) after successfultransmission is computed based on Equation 3:

f

RNCWNCW

wiaverage

newnew 1][][ 1

Where ( ][NCWnew ) is the new computedcontention window for a station ( N ),( ][1 NCWnew ) stands for the previous computedCW for a station ( N ), and the notation ( f ) standsfor a scaling factor (the impact of this factor isdiscussed later in this paper). Hence, the selectedCW size by a station ( N ) is obtained usingEquation.4. Equation 4 also guarantees that the( ][NCWnew ) size does not go below the minimumcontention window of a station ( N )(i.e. ][min NCW ).

][NCW = ])[],[( min NCWNCWMax new

3.2 Ratio Based Scheme (Case of Collision)In the legacy IEEE 802.11 DCF [4], after eachcollision, the CW is doubled. If the maximum limitknown as maximum Contention Window (CWmax) isreached, the collided station remains at CWmax. Inthe Ratio based scheme, after each collision, thenew CW of the collided station is computed asdepicted in Equation 5:

wiaveragenewnew RfNCWNCW 1][][ 1

Where ( ][NCWnew ), ( f ), ( ][1 NCWnew ) and

( wiaverageR ) are as discussed in the successful

transmission case. The selected CW value by astation ( N ) is obtained using Equation 6. Note that( ][max NCW ) is the maximum contention windowfor a station ( N ).Equation 6 also ensures that the( ][NCWnew ) size does not exceed the maximumcontention window (i.e. ][max NCW ).

][NCW = )],[( max newCWNCWMin

3.3 Collision Rate Variation SchemeThe Collision Rate Variation (CRV) scheme isbased on the variation in the collision ratio that wasdiscussed in the Ratio based scheme. Therefore, thisscheme is introduced in order to obtain furtherknowledge about the changes in the networkconditions by monitoring the variations in thecurrent and previous values of collision ratio. In theCRV scheme, the CRV value of each station iscalculated based on Equation 7

][][][ __ NRNRNCRV averagepreviousaveragecurrent

Where, ( ][NCRV ) is the collision ratio variationfor a station ( N ), ( ][_ NR averagecurrent ) and

( ][_ NR averageprevious ) stand for the current and theprevious average collision ratio for a station ( N ),respectively.

According to Equation 7, the CRV values arevaried between -1 to 1. The variation of CRV valuesis used to adjust the CW size and the DIFS for eachstation. This was discussed for each individualparameter in the following two sections.

Contention Window Adjustment UsingCollision Rate Variation Scheme.

In this section the operation of CRV scheme toadjust the CW size for each individual station basedon the variation in the CRV value is explained.Using the CRV scheme, the new CW size(i.e. ][NCWnew ) is updated using Equation 8.

][][][][ 11 NCRVNCWfNCWNCW newnewnew

(3)

(5)

(6)

(7)

(4)

(8)

(7)



Where, ( f ) refers to a scaling factor (discussedlater for more details), ( ][NCWnew ) is thecomputed CW for a station ( N ), ( ][1 NCWnew )stands for the previous CW size for a station ( N ),and ( ][NCRV ) is the computed collision ratiovariation for a station ( N ).

If the computed value of CRV using Equation 7 isnegative, this implies that the current number ofcollisions is less than previous number of collisions,therefore, the new CW size (i.e. ][NCWnew ) is usedby a station ( N ) after each successful transmissionand called as ][NCWsuccess . To ensure that the newCW for a station ( N ) (i.e. ][NCWnew ) does not gobelow the minimum contention window of a station( N ) (i.e. ][min NCW ), the CW size of the successfulstation ( N ) is limited by Equation 9.

])[,][(][ min NCWNCWMaxNCW new

If the computed CRV value is positive, thisimplies that the current number of collisions for astation ( N ) is more than the previous number ofcollisions for the same station. As a result, the newCW size (i.e. ][NCWnew ) is used by a station ( N )after each collision and called as ][NCWcollision . Toensure that the CW for a station ( N )(i.e. ][NCWnew ) does not exceed the maximumcontention window size of a station ( N ) (i.e.

][max NCW ), the CW size for a collided station( N ) is limited by using Equation 10.

])[,][(][ max NCWNCWMinNCW new

In the Ratio based and CRV schemes, the CWvalue does not reset to CWmin after successfultransmission except in the following cases: (1) atthe beginning of the transmission where the stationstarts with its initial CWmin. (2) When the stationexperiences a large CW size (i.e.,

][)1(][ min NCWfNCW ), in order to avoidstarvation. (3) When the number of retransmissionattempts of the collided packets reaches themaximum limit (i.e. station experiences high valueof CW).

In CRV scheme, the CW value is adjustedaccording to the variations in the collision ratio. Anegative CRV value determines that the stationexperiences less collision, indicating a more

successful transmission takes place while a positivevalue determines that the station experiences morecollisions. Therefore, the use of the CRV schemeprovides a good guide for selecting the CW value(i.e. successCW or collisionCW ). The appendixcontains: Ratio based scheme when collision occurs

(figure 1). Ratio based scheme when successful

transmission occurs (figure 2). Collision Rate Variation (CRV) scheme for

adjusting CW size (figure 3).

4. Simulation ModelTo evaluate the validity of the Ratio based and CRVschemes and compare their performance with thebasic IEEE 802.11 DCF and EIED schemes in termsof QoS parameters, the NS-2 simulation packagewas used [33]. The network model used employsthe topology shown in Figure 4 for differentscenarios. The parameters used in these simulationswere based on the IEEE 802.11 networkconfigurations [34] and they are summarised inTable 1.

Parameter ValueDIFS 50 µsecSIFS 10 µsecCWmin 31slotsCWmax 1023 slotsSlot Time 20 µsecUDP header 8 bytesIP header 20 bytesMAC header 28 bytesPHY header 24 bytesData Rate 2.0 MbpsRouting Protocol AODVIFQ Size 50

(9)

(10)

Table 1. IEEE 802.11 simulation settings [33] and[4].



Fig. 4. Single hop topology.

In all simulations, a Constant Bt Rate (CBR)traffic sources were employed. The packet sizesused for CBR traffic were 512, 160 and 200 bytes.The packet generation rates were 384 Kbps, 64Kbps and 128 Kbps.

The simulations were performed for severalscenarios in order to critically evaluate theperformance of the proposed schemes by means ofcomparison with the IEEE 802.11 DCF and theExponential Increase Exponential Decrease (EIED)schemes. These scenarios include varying thenetwork size (small, medium, and large networksizes), traffic load (light, medium, and heavy load),and the variation of the number of active stationover time.

5. Results and DiscussionsThis section provides the simulation results for theproposed methods and compares them with both,the Exponential Increase Exponential Decrease(EIED) with ri and rd factors equal 2 and theoriginal IEEE 802.11 DCF schemes.

5.1 Parameter TuningIn this section, the influence of history window size( wi ), scaling factor ( f ) and weighting factor ( )is investigated. The results showed that the right setof these parameters can lead to better performance.In order to ensure that these parameters wereappropriately selected, several simulations were

carried out with different topologies and differenttraffic loads. In this discussion, a network with 10stations transmitted CBR traffic to 10 destinations inheavy and medium load cases. Average throughputand average delay were considered as the mainmetrics.

The weighting factor ( ) was varied in the rangeof 0 to 1. Figures 5a and 5b depict the average delayand average throughput as a function of weightingfactor ( ), respectively. Every single point of theresults obtained represented an average of 10simulations in order to avoid the bias of randomnumber generation. It can be observed from Figure5a that the lowest average delay corresponded to = 0.65. Figure 5b shows that the highest averagethroughput corresponded to = 0.7. Consequently,the value of equal 0.6 was selected since itachieved a tradeoff between average throughput andaverage delay.

Fig. 5. Average delay and average throughput as afunction of weighting factor ( ), (a) average delayand (b) Average throughput.

600

650

700

750

800

850

900

950

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Weighting factor (λ)

Ave

rage

del

ay (m

sec)

1200

1215

1230

1245

1260

1275

1290

1305

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1Weighting factor (λ)

Aver

age

thro

ughp

ut (K

bps)

(a)

(b)



0

200

400

600

800

1000

1200

1400

1600

1 2 3 4 5 6 7 8 9 10

Scalling factor (f)

Avera

ge de

lay (m

sec)

heavy load trafficmedium load traffic

1000

1050

1100

1150

1200

1250

1300

1350

1400

1 2 3 4 5 6 7 8 9 10

Scalling factor (f)

Aver

age t

hrou

ghpu

t (kb

ps)


Figures 6a and 6b show the average delay andaverage throughput as a function of scaling factor( f ) in heavy and medium load cases. The scalingfactor ( f ) was varied in the range of 1 to 10 withan increment of 1. A significantly low value of ( f ),e.g. f =1, resulted in high values of delay and lowvalues of average throughput in heavy and mediumload cases. Similarly, a significantly high value of

( f ), e.g. f = 10, resulted also in high values ofaverage delay and low value of average throughput.According to Figures 6a and 6b, an appropriatevalue of ( f ), i.e. around 3, provided a tradeoffbetween the average delay and average throughput.The value of ( f ), i.e. around 3, provided the lowestaverage delay and the highest average throughput inheavy and medium load cases as depicted in Figures6a and 6b. It can be observed that the variation inthe value of the scaling factor ( f ) could result in asignificant variation in the network performance.Therefore, this feature will be used with otherparameters such as packet loss rate for providingservice differentiation in single and multi-hopnetworks. In this study, the value of f = 3 wasconsidered for the following simulations.

Figures 7a and 7b demonstrate the effect of thehistory window ( wi ) on the average delay andaverage throughput in heavy and medium loadcases. The size of history window ( wi ) was variedin the range of 5 to 50 with an increment of 5packets. The simulation results of average delay areshown in Figures 7a for heavy and medium loadcases. The value of ( wi ) around 20 provided thelowest values of average delay in heavy andmedium load cases. According to Figure 7b, thehighest average throughput was achieved when the( wi ) size was around 25 in heavy load case, whilethe highest average throughput was achieved whenthe ( wi ) size was around 20 in medium load case.Therefore, the history window ( wi ) around 20 wasselected for the following simulations since itprovided a balance between the heavy and mediumloaded cases. Furthermore, it provided a tradeoffbetween the average delay and average throughput.

Note that, the history window ( wi ) valuesaround 20 could maintain small values of averagedelay compared to smaller values of ( wi ).Therefore, a history window size equal 20 was usedin this chapter as indicated in the previousparagraph. Because too small values of ( wi )probably was insufficient to provide adequateinformation about the current and the previousnetwork conditions. Moreover, too large values of( wi ) could cause late update in the adaptationprocess (i.e. adjusting CW and DIFS) in whichperformance degradation might occur.Subsequently, choosing an appropriate size of ( wi )result in minimum average delay and maximumaverage throughput.

(a)

(b)

Fig. 6. Average delay and average throughput asa function of scaling factor ( f ), (a) averagedelay and (b) Average throughput.



200

300

400

500

600

700

800

900

1000

1100

5 10 15 20 25 30 35 40 45 50History window wi (packet)

Aver

age d

elay (

mse

c)


1150

1160

1170

1180

1190

1200

1210

1220

1230

1240

5 10 15 20 25 30 35 40 45 50History of sent packets wi (packet)

Aver

age

thro

ughp

ut (K

bps)


Fig. 7. Average delay and average throughput as afunction of history window ( wi ), (a) average delayand (b) average throughput .

5.2 Performance Evaluation of CW AdjustmentsWith Heavy Load TrafficThe Ratio based and CRV schemes were validatedfor heavy load traffic. Different network topologiessizes (5, 10 and 20 connections, i.e. small, mediumand large networks) were considered for thispurpose. Additionally, some other scenarios wereintroduced in order to critically investigate thebehavior of the Ratio based and CRV schemes. Thisincluded the case where the sources transmitteddifferent traffic types.

The offered load delivered into the network forthe heavy load traffic was approximately 80% of thechannel capacity. This implied that 1.6 Mbps were

transmitted by all the active senders in the network.Therefore, the transmission rate of each source wasequal ( Mbps6.1 / n ), where n represents thenumber of connections.

Small Network Size (5 Connections). Thissimulation consisted of 40 stations as shown inFigure 4; only 5 connections transmitted at heavyload to 5 destinations. Around 80% of the channelcapacity (more than 1600 Kbps) was offered to thenetwork. The transmission rate of each source was320 Kbps CBR traffic.

Figure 8a shows the average delay for the fourschemes. The Ratio based and the CRV schemeswere able to maintain low average delay at heavilyloaded traffic. It can be seen that the average delaywas 57% and 50% smaller than that for the legacyIEEE 802.11 DCF and the EIED schemerespectively when the Ratio based scheme wasemployed. Similarly, the CRV scheme alsooutperformed the IEEE 802.11 DCF and EIEDschemes by 55% and 48% respectively. Indeed, thetwo proposed schemes provide a lower delay (lessthan 400 msec). The CRV scheme showed morevariations especially at the beginning and at the endof simulation. This was due to the lack of networkhistory since the number of active stations in thenetwork was very small and each station transmittedlarge number of packets which required severaladjustments of successCW and collisionCW . Once thesystem selected proper values of successCW and

collisionCW these fluctuations became less.Furthermore, the fluctuations became less when thenumber of active stations increased as discussedlater.

As indicated in Table 2, the Ratio based and CRVschemes achieved small values of average jitter lessthan 10 msec. The improvements reached 22% and15% when the Ratio based scheme was used andreached 41% and 36% when the CRV scheme wasemployed over the IEEE 802.11 DCF and EIEDschemes respectively. Moreover, the CRV had amean jitter 24% less than the Ratio based scheme.High value of jitter in the IEEE 802.11 DCF wascaused by the sharp variations in the CW size,which resulted in a large variation in the BackoffInterval (BI), and consequently led to large valuesof delay and jitter.

(b)

(a)

Collision RateVariation (CRV)schemeforadjusting CWsize

[wi]= 20,flag = 0;f = 3,CWmin =



0

100

200

300

400

500

600

700

800

900

0 30 60 90 120 150 180 210 240 270 300

Simulation time (sec)

Aver

age

dela

y (m

sec)

2

4

6

8

10

12

14

16

0 30 60 90 120 150 180 210 240 270 300


Ave

rage

col

lisio

n (%

)

0

20

0 20

Ratio basedIEEE 802.11DCFExponential Increase Exponential Decrease (EIED)Collision Rate Variation (CRV)

Fig. 8.The network parameters for 5 connections atheavy CBR traffic, (a) average delay (msec) and (b)average collision rate (%).

As discussed earlier, the MAC efficiency and thecollision rate were related to each other. Thisimplied that the reduction in the number ofcollisions led to an improvement in the performanceof the protocol (i.e. improve the MAC efficiency byincreasing the number of successful packetstransmission with respect to the total number ofpackets sent). In term of collision rate, the CRVscheme had the superiority over other schemes. Thecollision rate obtained was considerably lower thanthe values for IEEE 802.11 DCF and EIED schemesas shown in Figure 8b.

The CRV scheme had the best performance. Ithad less delay and less collision than the IEEE802.11 DCF, EIED and the Ratio based schemes,respectively. The quantitative statistics for a small

network scenario in heavy load network aresummarized in Table 2.

Table 2. Statistical results obtained for fourdifferent schemes in a small size network (5connections) at heavy load CBR traffic.

ParameterRatio

BasedScheme

IEEE

802.11DCFScheme

EIEDscheme

CRVscheme

Average delay(msec) 345 810 699 365

Average jitter(msec) 9.9 12.8 11.8 7.6

Averagethroughput(Kbps) 1297 1227 1252 1227

Average packetloss (%) 7.8 21.1 16.4 9.0

Average MACEfficiency (%) 92.9 85.7 86.8 95

Averagecollision (%) 7.0 13.9 11.9 4.9

Medium Network Size (10 Connections). Inthis scenario, the topology shown in Figure 4 wasused. The offered load was 80% of the channelcapacity which was equally distributed among the10 connections. Each source transmitted 160 Kbpsto its corresponding destination.

The values of average delay and jitter for allschemes were increased when the number of activestations was increased from 5 connections to 10connections with the same a mount of traffic. Forinstance, the values of average delay were increasedby 44%, 55%, 54% and 47% in the Ratio based,IEEE 802.11 DCF, EIED, and CRV schemesrespectively. This implied that the contentionbetween stations became significant factor. InFigure 9a, the Ratio based and the CRV schemesshowed a smaller mean delay. For instance, areduction of 65% and 59% was observed in theaverage delay when the Ratio based scheme wasemployed compared to the IEEE 802.11 DCF andEIED schemes respectively. Similarly, the CRVscheme achieved a small average delay which was9% higher than that obtained for the Ratio basedscheme. Note that in case of 5 connections, theRatio based and the CRV schemes experienced highfluctuations. In case of 10 connections, these

(a)

(b)



fluctuations became less, because of the properselection of the CW size when the network becamebusier.

According to figure 9b high values of collisionrate were observed especially at the first 30 secondsof the simulation. This was due to the impact ofrouting information exchange during the initialperiod of the simulation which, once established,this effect became less along with the simulationtime. The Ratio based scheme achieved betterperformance compared with other schemes. Thequantitative results of the assessed QoS and otherQoS parameters for medium size network aresummarised in Table 3.

Table 3. Statistical results obtained for fourdifferent schemes in a medium size network (10connections) at heavy load CBR traffic

300

600

900

1200

1500

1800

2100

0 30 60 90 120 150 180 210 240 270 300


Aver

age d

elay (

mse

c)

0

20

0 20

R a tio ba s e dIE E E 8 0 2 .1 1 D C FE x po ne ntia l Incre a s e E x po ne ntia l D e cre a s e (E IE D )C o llis io n R a te V a ria tio n (C R V )

4

6

8

10

12

14

16

18

20

0 30 60 90 120 150 180 210 240 270 300Simulation time (sec)

Aver

age c

ollisi

on (%

)

Fig. 9.The network parameters for 10 connectionsat heavy CBR traffic, (a) average delay (msec) and(b) average collision rate (%).

Large Network Size (10 Connections). Theperformance of the four schemes was negativelyaffected when the network size was increased froma small network size (i.e., 5 connections) to amedium network size (i.e., 10 connections).However, this impact could be significant as thenetwork size became larger. In this section, theperformance of the four schemes is evaluated whenthe number of active stations was increased to 20connections. The volume of CBR traffic represented80% of channel capacity where each sourcetransmitted 80 Kbps.

ParameterRatio

BasedScheme

IEEE

802.11DCFScheme

EIEDscheme

CRVscheme

Average delay (msec) 621.6 1788.0 1520 681.9

Average jitter (msec) 21.3 40.1 36.3 16.5

Average throughput(Kbps)

1240 1025.5 1089 1160

Average packet loss(%) 7.9 27.1 19.5 9.6

Average MACefficiency (%) 91.4 81.1 83.7 92.7

Average collision (%) 8.6 17.0 14.9 6.3

(a)Fig. 2.Ratiobasedschemein caseofsuccessfultransmission

(b)



Figures 10a and 10b show that the performanceof the four schemes was degraded in 20 connectionsnetwork. However, the Ratio based and CRVschemes still performed better than the other twoschemes. The mean delay was 57% less than theobtained delay when the IEEE 802.11 DCF andEIED schemes were used. The mean jitter value forthe Ratio based scheme were 47% less than thevalues obtained when the IEEE 802.11 DCF andEIED schemes were employed. Similarly, theaverage throughput improved by 19% and 15%compared to the IEEE 802.11 DCF and EIEDschemes respectively. This was due to the reductionin the number of packets drop, where only less than11% was lost when the Ratio based and CRVschemes were used. On the other hand, more than24% of packets were lost when the IEEE 802.11DCF and EIED schemes were used.

300

800

1300

1800

2300

2800

3300

3800

4300

0 30 60 90 120 150 180 210 240 270 300


Aver

age d

elay (

mse

c)

10

11

12

13

14

15

16

17

18

19

20

0 30 60 90 120 150 180 210 240 270 300


Aver

age c

ollisi

on (%

)As indicated in Table 4, the IEEE 802.11 DCF

had a mean MAC efficiency of 78% which was11% less than the one obtained for the Ratio basedscheme. The other schemes also displayedimprovements in their MAC efficiency values. Thisimplied that each scheme was able to adjust itsbackoff timer, particularly the CW size, until thebehaviour of each scheme became stable. However,the Ratio based and the CRV schemes achievedMAC efficiency 10% higher than the other twoschemes. Because the Ratio based and CRVschemes considered part of the network history totune their backoff timer by getting a suitable CWsize after successful and unsuccessful transmission;whereas the other two schemes only considered thecurrent network state.

The collision rate obtained is shown in Figure10b. The collision rates achieved by the fourschemes were similar when the traffic load waslight. However, at heavy load traffic, the Ratiobased and the CRV schemes were able to maintain alower collision rate than the IEEE 802.11 DCF andEIED schemes. This behaviour can be explained bythe fact that the Ratio based and CRV schemes usedan adaptive mechanism to adjust the CW size basedon the collision rate history the stationsexperienced. As a result, a considerable reduction inthe collision rate values was obtained whichimproved the network performance. The statisticalresults for this scenario are summarised in Table 4.

Table 4. Statistical results obtained for fourdifferent schemes in a large size network (30connections) at heavy load CBR traffic.

ParameterRatiobasedscheme

IEEE802.11DCFscheme

EIEDscheme

CRVscheme

Average Delay (msec) 1465 3348 3635 1180

Average jitter (msec) 48.2 95.2 92.7 33.9

Average Throughput(Kbps) 1147 928 972.0 974.4

Average loss (%) 11.0 24.7 23.9 8.0

Average MAC efficiency (%) 88.7 78.7 80.4 88.1

Average collision rate (%) 11.3 17.2 16.9 11.2

Fig. 10.The network parameters for 10connections at heavy CBR traffic, (a) averagedelay (msec) and (b) average collision rate (%).

(a)

(b)



It is worth noting that, the trend of the curves forall schemes was smoother when the network sizewas increased. This was due to the reduction in thenumber of packets sent by each station whichrequired smaller adjustments of the CW size foreach station.

The network performance was affected when thenumber of transmitting stations was increased. Thisimplied that the network size and the offered loadplayed a major role in the performance of ad-hocnetworks. Figures 11 depicts the collision rate as afunction of the network size for the four schemes.

In Figure 11, the performance of a small network(i.e. 5 connections) was better than the medium andlarge ones. As expected, delay, jitter and packet lossof all schemes increased with the increase in thenetwork size because of large CW values (highcompetition between contending stations).Furthermore, the network with 20 connectionscaused large number of collisions due to the highcompetition which also led to less MAC efficiency.However, the proposed schemes achieved betterperformance than the two other schemes, whateverthe number of connections is. For instance, at heavyload case for 5, 10 and 20 connections, the IEEE802.11 DCF and EIED achieved poor performance,while the Ratio based and the CRV schemes, theymaintained higher performance for all networksizes.

02468

1012141618

heavy load heavy load heavy load

5 connections 10 connections 20 connectionsNetwork size

Ave

rage

colli

sion

(%)

Ratio basedIEEE 802.11 DCFExponential Increase Exponential Decrease (EIED)Collision Rate Variation (CRV)

Fig. 11. Collision rate as a function of network sizeat heavy load CBR traffic.

Ratio Based and Collision Rate Variation in aRealistic Scenario

The performance of the Ratio based and CRVschemes was evaluated against an increasingnumber of active stations over time. This wascarried out in order to examine the performance ofthese schemes when the network was experiencedhighly changing configurations. Here, the networktopology shown in Figure 4 was used with twentystations transmitting CBR traffic, using a 512 bytespacket size, to twenty different destinations usingthe basic access mechanism. The simulation timewas 400 seconds. Every 5 seconds a new CBRsource with 80Kbps generation rate started itstransmission. At the 100th second of the simulation,20 sources were active in the network (i.e.contending to access the channel) and sending videopackets to 20 destinations. These 20 CBR sourcesremained active to the 300th second in order tosustain heavy load throughout the 200 seconds (i.e.from 100 to 300 seconds). At the 300th second, thenumber of active stations was reduced by one every5 seconds until all sources stopped theirtransmission at the 400th second.

According to Figure 12, the value of averagedelay increased with the simulation time (due to theincrease of the number of active stations). However,the Ratio based and CRV schemes maintained 50%of average delay when the IEEE 802.11 DCF andEIED schemes were employed. The maximumvalues of average delay were observed between the200th and 300th second because of highcompetition between the contending stations. In theIEEE 802.11 DCF and EIED schemes each stationselected a large CW size in order to avoid collisionsat the cost of wasting several idle time slots whichin turn led to high values of delay. The sharptransition from a large CW size to CWmin in case ofIEEE 802.11 DCF and to half of the current CWsize of the EIED scheme after successfultransmission increased the amount of jitter.Thereafter, the average delay started to decreasesince the number of sources decreased by one every5 seconds.

The Ratio based and CRV schemes maintainedapproximately similar values of delay as the othertwo schemes up to 100th second, since the networkwas still lightly loaded and the number ofcontending stations was less than 20 connections.After the 100th second, the network became busierand the load heavier, therefore, the Ratio based andCRV schemes performed better and maintainedlower values of average delay and average jitter.This was due to the capability of the Ratio based



and CRV of adaptively selecting the CW size aftersuccessful and unsuccessful transmission in a waythat achieved a tradeoff between collisions decreaseand idle time slots increase.

Similarly, the average throughput was 16% and11% higher than the IEEE 802.11 DCF and EIEDschemes, respectively when the Ratio based schemewas used and 11.6% and 6% higher when the CRVscheme was employed. The reduction in the averagethroughput and the increase in the number of packetloss when the IEEE 802.11 DCF and EIED schemeswere used were due to the following; in the IEEE802.11 DCF scheme, as the station resets its CW toCWmin after successful transmission or decreases itto the half of its current value in EIED scheme, thestation forgets about the collision history. In thiscase when all stations kept transmitting with thesame data rate; it is likely that the new transmissionnoticed contention and collisions as before. This inturn increased the collision rate especially during ahigh contention period as shown in Figures 12b.This was mitigated by keeping some history of theobserved successful and collisions packets. In thiscase instead of resetting the CW to CWmin aftersuccessful transmission or doubling it aftercollisions, the CW size was changed adaptivelybased on the history of collision rate.

The behaviour of the Ratio based and CRVschemes was apparent on both the achieved MACefficiency and the collision rate parameters. Forinstance, the Ratio based scheme had a 90% averageMAC efficiency and 8.3% average collision rate;whereas 81% average MAC efficiency and 14%average collision rate were observed for the IEEE802.11 DCF scheme. The CRV scheme alsoachieved higher performance than the IEEE 802.11DCF and EIED schemes.

It can be concluded that the Ratio based and CRVschemes were capable to adaptively adjust the CWvalues after successful and unsuccessfultransmission based on the history of each individualstation. Moreover, both schemes were able toachieve an efficient tradeoff between collisiondecrease and idle time slots increase in a way thatQoS for the transmitted application was achieved.Since the original protocol and the EIEDmechanism cause a sharp decrease or a sharpincrease of the CW value after successful orunsuccessful transmission. Additionally, bothschemes do not consider the history of the networkcondition as the proposed schemes.

Fig 12. QoS parameters when the number ofconnections was changed over time, (a) averagedelay (msec), and (b) average collision rate (%)

6. ConclusionsThe main objective of this paper is to describe thedeveloped adaptive techniques namely Ratio basedand CRV scheme that were used to enhance theperformance and to improve the QoS of the IEEE802.11 DCF scheme.

The Ratio based and CRV schemes extended thelegacy IEEE 802.11 DCF mechanism bydynamically adjusting the CW value for each stationaccording to the current and past history ofsuccessful and unsuccessful packet transmissions.The aim of developing these approaches was toreduce the probability of collisions in an attempt toimprove QoS in IEEE 802.11 DCF protocol. The



Ratio based and CRV schemes are easy toimplement since they do not require majormodifications to the IEEE 802.11 DCF framesformat. The simulation results indicated that theRatio based and CRV showed better performancethan the other two schemes regardless of thenetwork size, traffic type, and the access mechanismused. For instance, the average delay reduced by59% and 56% as compared with the standard IEEE802.11 DCF and EIED schemes, respectively. TheCRV scheme performed better than the Ratio basedthe IEEE 802.11 DCF, and the EIED schemes inmost scenarios.

References:[1] Bray J., Sturman Ch. F., and Mendolia, J.,

Bluetooth 1.1 Connect without cables, PrenticeHall, 2001.

[2] Ilyas M. and Qazi S., Ad hoc WirelessNetworks-Challenges and Opportunities, inInternational Workshop Frontiers, FIT03,2003.

[3] Tsai T. and Tu C., An Adaptive IEEE 802.11MAC in Multihop Wireless Ad Hoc NetworksConsidering Large Interference Range, inProceedings of WONS2004, Wireless On-Demand Network Systems, Page(s): 86-99,2004.

[4] IEEE Standard for Wireless LAN MediumAccess Control (MAC) and Physical Layer(PHY) Specifications, ISO/IEC 8802-11:1999E), 1999.

[5] Crow B., Widjaja I., Kim L., and Sakai P.,IEEE 802.11 Wireless Local Area Networks, inIEEE Communications Magazine, vol. 35, no.9, Page(s): 116-126, 1997.

[6] Le L. and Rhee I., Implementation andExperimental Evaluation of Multi ChannelMac Protocols for 802.11 Networks, Ad HocAd Hoc Networks Volume 8, Issue 6, Pages626-639, 2010.

[7] Hung F., Marsic I., Performance analysis ofthe IEEE 802.11 DCF in the presence of thehidden stations , Computer Networks, Volume54, Issue 15, 28, Pages 2674-2687, 2010.

[8] Baldwin R., Davis N., and Midkiff S., A Real-Time Medium Access Control Protocol for AdHoc Wireless Local Area Networks, in ACMMobile Computing and Communication, vol.3, no. 2, 1999.

[9] Sobrinho J. and Krishnakumar A., Quality-of-Service in Ad Hoc Carrier Sense Multiple

Access Wireless Networks, in IEEE Journal,Selected Areas in Communications, vol. 17,no. 8, Page(s): 1353 - 1368, 1999.

[10] Cali F., Conti M., and Gregori E., DynamicTuning of the IEEE 802.11 DCF 802.11Protocol to Achieve a Theoretical ThroughputLimit, in IEEE/ ACM Transaction Networking,vol. 8, no. 6, Page(s): 785 - 799, 2000.

[11] Bianchi G. and Tinnirello I., :Kalman FilterEstimation of the Number of CompetingTerminals in an IEEE 802.11 Network, inProceeding IEEE INFOCOM ’03, 2003.

[12] Qixiang P., Soung L., Jack Y., Gary C., ATCP- Like Adaptive Contention WindowScheme for WLAN, in IEEE InternationalConference on Communications, vol. 6,Page(s): 3723-3727, 2004.

[13] Kuo W. and Jay C., Enhanced Backoff Schemein CSMA/CA for IEEE 802.11, in the IEEEVehicular Technology Conference, 2003.

[14] Kwon Y., Fang Y., and Latchman H., A NovelMAC Protocol with Fast Collision Resolutionfor Wireless LANs, in Proceeding IEEEINFOCOM ’03, 2003.

[15] Zhao J., Guo Z., Zhang Q., and W. Zhu,Distributed MAC Adaptation for WLAN QoSDifferentiation, in Global TelecommunicationsConference, vol. 6, Page(s): 3442-3446, 2003.

[16] Gannoune L., Robert S., Dynamic Tuning ofthe Contention Window Minimum (CWmin)for Enhanced Service Differentiation in IEEE802.11 Wireless Ad-hoc Networks, in IEEEInternational Symposium on Personal, Indoorand Mobile Radio Communications(PIMRC'04), 2004.

[17] Deng J., Varshney P., and Haas Z., A NewBackoff Algorithm for the IEEE 802.11Distributed Coordination Function, inProceeding of Communication Networks andDistributed Systems Modeling andSimulation(CNDS), 2004.

[18] Kuppa S. and Prakash R., Adaptive IEEE802.11 DCF Scheme with Knowledge-BasedBackoff, in Proceeding IEEE WCNC, 2005.

[19] Robinson J. and Randhawa T., SaturationThroughput Analysis of IEEE 802.11eEnhanced Distributed Coordination Function,in IEEE JSAC, vol. 22, Page(s): 917-928,2004.

[20] Ksentini A, Naimi M., Nafaa A., and GuerouiM., :Adaptive Service Differentiation for QoSProvisioning in IEEE 802.11 Wireless ad hocNetworks, in PE-WASUN, Page(s): 39-45,2004.



[21] Zhang S. and Ye C., On serviceDifferentiation in Mobile Ad hoc Networks, J.Zhejiang University Science, vol.5, no.9,Page(s): 1087-1094, 2004.

[22] Pattara-atikom W., Banerjee S., andKrishnamurthy P., A-DRAFT: An AdaptiveQoS Mechanism to Support Absolute andRelative Throughput in 802.11 Wireless LANs,in Proceeding ACM International Workshopon Modeling Analysis and Simulation ofWireless and Mobile Systems (MSWiM),Page(s): 117 - 126, 2004.

[23] Sung M. and Yun N., A MAC ParameterOptimization Scheme for IEEE 802.11e-basedMultimedia Home Networks, in the IEEECCNC 2006 proceedings, 2006.

[24] Bianchi G., Fratta L., Oliveri M., PerformanceEvaluation and Enhancement of theCSMA/CA MAC Protocol for 802.11 WirelessLANs, in Proceeding, PIMRC’96, Page(s):392–396, 1996.

[25] Cali F., Conti M., and Gregori E., IEEE 802.11Protocol: Design and Performance Evaluationof an Adaptive Backoff Mechanism, in IEEEJournal Selected Areas in Communication, vol.18, no. 9, Page(s): 1774–1786, 2000.

[26] Qiao D. and Shin K., UMAV: A SimpleEnhancement to the IEEE 802.11 DCF, inProceeding the 36th Hawaii InternationalConference on System Sciences (HICSS-36),2003.

[27] Choi J., Yoo J., Choi S, and Kim C, EBA: anEnhancement of the IEEE 802.11 DCF viaDistributed Reservation Mobile Computing, inIEEE Transactions, vol. 4, no. 4, Page(s): 378 -390, 2005.

[28] Zizzo, G., Sanseverino E., Silvestre M.,Campoccia F, A modified IEEE 802.11protocol for reliable data transmission inautomated power distribution systems, InProceedings of the 44th InternationalUniversities Power Engineering Conference,2009.

[29] Puthal, D., Quality of Service Provisioningwith modified IEEE 802.11 MAC Protocol.MTech by Research thesis, 2008.

[30] Bensaou, B., Fang Z., A Fair MAC Protocolfor IEEE 802.11-Based Ad Hoc Networks:Design and Implementation. In IEEETransactions on Wireless Communications,Volume: 6, Issue: 8, Pages: 2934-2941, 2007.

[31] Song N., Kwak B., Song J., and Miller L.,Enhancement of IEEE 802.11 DistributedCoordination Function with Exponential

Increase Exponential Decrease BackoffAlgorithm, in IEEE VTC 2003, Page(s): 2775-2778, 2003.

[32] Crowder S., Design of ExponentiallyWeighted Moving Average Schemes in Journalof Quality Technology, vol. 21, no. 3, Page(s):155-162, 1989.

[33] NS-2, [online]. Last accessed on 6 February2009 at: http://www.isi.edu/nsnam/ns.

[34] Choi S., :PCF vs. DCF: Limitations andTrends, in IEEE 802.11-01/154, 2001.

[35] ETSI TS 101 761-1, Broadband Radio AccessNetworks (BRAN), HIPERLAN Type 2", ETSI,2000.

[36] Bharghavan V., Demers A., Shenker S., andZhang L., :MACAW: a Media Access Protocolfor Wireless LAN’s, in Proceeding of theconference on Communications architectures,protocols and applications, Page(s): 212–225,1994.

[37] R Nair, B Rajagopalan, H Sandick,Aframework for QoS- based Routing, ECrawley, in The Internet Engineering TaskForce, 1998.

[38] Caprihan G, Kumar S, Saran H, A QoStranslation methodology based on userperception. IEEE Singapore InternationalConference on Networking, Singapore, 1997.



Appendix

Fig. 1. Ratio based scheme in case of unsuccessful transmission.

Ratio based scheme when collision occurs [wi] = 20, flag = 0; f = 3, CWmin = 31; CWmax =1023; If (history window == [wi] packets) {

Count the number of collided packets from wiCount the number of successfully received acknowledgment packets from wiCompute the current collision ratio using Eq. 1;Compute the average collision ratio using Eq. 2;

//to avoid starvation for some stations monitor the behavior of each I individual station.If (CW size is grater than > (f +1) * CWmin ) {

Increment flag;I f( flag == f +1) {

CW = CW min ; } else {

flage = 0;Compute CW size using Eq .5;

Apply Equation 6; }

} }

Ratio based scheme when successful transmission occurs [wi] = 20, flag = 0; f = 3, CWmin = 31; CWmax =1023; If ( history window == [wi] packets) { Count the number of collided packets from wi; Count the number of successfully received acknowledgment packets from wi;

Compute the current collision ratio using Eq. 1;Compute the average collision ratio using Eq. .2;Reset the collision counter;Reset the success counter;//to avoid starvation for some stations monitor the behavior of each individual station.If (CW size is grater than > (f +1) * CWmin ) {


CW = CW min ; } else {

flage = 0;Compute CW size using Eq. .3;

Apply Equation 4; }

} }

Fig. 2. Ratio based scheme in case of successful transmission.



Fig. 3. Collision Rate Variation scheme (CRV).

Collision Rate Variation (CRV) scheme for adjusting CW size [wi] = 20, flag = 0; f = 3, CWmin = 31; CWmax =1023; CRV =0 ; If (history window = = [wi] packets) {

Count the number of collided packets from wi; Count the number of successfully received

acknowledgment packets from wi;Compute the current collision ratio using Eq. 1;Compute the average collision ratio using Eq. 2;Reset the collision counter;Reset the success counter;

Compute the collision rate variation value for each station using Eq. 7;If (CRV[N] < 0) {

//to avoid starvation for some stations monitor the behavior of each station.If (CW size is grater than > (f +1) * CWmin ) {

Increment flag; If( flag = = f +1) {

CW = CW min ; } else {flage = 0;Compute CW size using Eq. 8;Apply Eq. 9;

Use the computed CW size after successful transmission as Follows: CWsuccess[N] = CW[N]; }

}}else If (CRV[N] > 0) {

//to avoid starvation for some stations monitor the behavior of each stationIf (CW size is grater than > (f +1) * CWmin ) {


CW = CW min ; } else {flage = 0;

Compute CW size using Eq. 8;Apply Eq. 10;Use the computed CW size after unsuccessful transmission as follow: CWcollision[N] = CW[N];

}}

} }



Documents

Novel Mechanisms for Quality of Service Improvements in ... · Novel Mechanisms for Quality of Service Improvements in Wireless Ad Hoc Networks ... nidal}@mutah.edu.jo Abstract: -