ebook img

Flexible and efficient aggregation framework for antifragile wireless mesh networks PDF

13 Pages·2015·1.88 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Flexible and efficient aggregation framework for antifragile wireless mesh networks

JReliableIntellEnviron(2015)1:159–171 DOI10.1007/s40860-015-0011-5 ORIGINAL ARTICLE Flexible and efficient aggregation framework for antifragile wireless mesh networks MuhammadShahzadAsif1 · MuhammadShafiq2 · Jin-GhooChoi2 · MuddesarIqbal3 · AzeemIrshad4 Received:16August2015/Accepted:24October2015/Publishedonline:5November2015 ©SpringerInternationalPublishingSwitzerland2015 Abstract Wirelessmeshnetworks(WMNs)havefoundan thatwhenandwhichtypeofaggregationissuitableinagiven interesting alternative application in public safety and dis- scenarioisacomplexbalancingact.Because,itisboundedby aster recovery as they are enabled with key features such variouslivestatisticsofthecommunicationlink(e.g.,buffer as fault tolerance, broadband support, and interoperability. size, link quality, bandwidth, maximum transmission unit, However, their sparsely distributed wireless nodes need to anddelayetc.)andthus,canconsiderablyaffectthenetwork frequently share the control packets among each other for efficiency.Ifsuchstatisticsdonotsupportaggregationorits successfuldatatransfer.Thesepacketsgiverisetoaconsid- certaintype,networkperformancemaycauseworseaffects. erableamountofcontroloverhead,especiallyformultimedia This paper proposes a decision-oriented dynamic solution, traffic,whichisnotbearableinjeopardysituationsofnetwork namely AADM, which is an adaptive aggregation-based disastersuchas9/11andHurricaneKatrina.Hence,toavoid decision model. Based on the link quality and its live sta- the huge cost of control overhead, aggregation is supposed tistics,AADMdynamicallytakesjudiciousdecisionsabout to be one of the handy solutions for building a new object aggregationanditsspecifictypetoachievethedesiredout- fromoneormoreexistingobjectsofnetworktraffic.Gener- come.Despitenetworkscalability,qualityofservice,power ally,aggregationhasthreetypestobeexecutedon‘packets’, optimization, and network efficiency, it also reduces con- ‘frames’andthe‘links’ofanetwork.Thenetworkdecision trol traffic in WMNs. OMNET++ simulations are used to verify AADM. Simulation results have shown that AADM outperforms existing static approaches in terms of packet B Jin-GhooChoi loss,throughputanddelay. [email protected] MuhammadShahzadAsif Keywords Disaster recovery · Packet aggregation · [email protected] Adaptiveaggregation·WMN·MTUandVoIP MuhammadShafiq shafi[email protected] MuddesarIqbal 1 Introduction [email protected] AzeemIrshad Disasters such as earthquakes and floods affect large geo- [email protected] graphicalareas,whichcreatesfragileenvironmentforrescue 1 UniversityInstituteofInformationTechnology,PMAS-Arid servicestooperate.Oneofthemajoraftermathsofthesedis- AgricultureUniversity,Rawalpindi,Pakistan aster situations is lack of communication infrastructure to 2 DepartmentofInformationandCommunicationEngineering, supportapplicationsthatcouldprovidecoordinationamong YeungnamUniversity,Gyeongsan,Korea alltherescueteamsanddisseminatenecessaryinformation 3 Pak-UKInstituteofInnovativeTechnologiesforDisaster and data for informed decision making. This requires an Management,UniversityofGujrat,Gujrat,Pakistan antifragileandresilienceapplicationframeworkthatcanalso 4 DepartmentofComputerScienceandSoftwareEngineering, provideon-demandcomputingresourcesandjudiciousdeci- InternationalIslamicUniversity,Islamabad,Pakistan sionmaking.Giventhisopportunity,wirelessmeshnetworks 123 160 JReliableIntellEnviron(2015)1:159–171 (WMNs) have found an interesting alternative application considerableamountofoverheadforcontrolinformationin in public safety and disaster recovery (PSDR) as they are wirelessnetworks[2].Moreover,suchincreasedoverheadis enabledwithkeyfeaturessuchasfaulttolerance,broadband also not affordable in jeopardy situations such as 9/11 and support,andinteroperability. HurricaneKatrina.Thispaperaddressesthestate-of-the-art TypicallyaWMNismadeupofradionodesorganizedin aggregationapproachesbringingaboutsomeimminentmod- ameshtopology.Whereby,aWMNisagroupofunplanned ificationsthatwouldnotonlyincreasethroughputbutcould nodesandmeshroutersinterconnectedthroughwirelesslinks alsousetoproducetheoptimumresultsinregardtonetwork [1]. The wireless service access point is installed at each performance. network user’s locale requiring minimal configuration and We proposed a flexible and efficient aggregation frame- infrastructure.InWMN,eachnetworkactsasanautonomous work for WMN named as adaptive aggregation based userandforwardsdatatraffictothenextnode.Thenetwork decision model (AADM) and is supported by OMNeT++ infrastructure is decentralized, in which these autonomous simulation.InAADMtheaggregationisperformed,basedon nodes send and receive data independently and is not con- runtimedecisionsafterevaluatingthemoredynamicfactors trolledbythecentralizedcontroller. likecongestion,routing,reliability,averagedelayandenergy A WMN usually consists of three types of nodes: mesh consumption.TheAADMexclusivelymonitorsthelivesta- clients (MC), mesh router (MR), and mesh gateway (MG). tistics of a network channel and takes run time decisions Clients can be either stationary or mobile, and can form a thateitheraggregationisneededornot?Ifneeded,whether mesh network among them by involving the mesh routers. it needs to aggregate on real time or non-real time basis? Whereas the wireless routers in WMN relay packets orig- Then it defines the maximum transmission unit (MTU) for inating from the client nodes to the gateway nodes, which aggregationdependingonaforementionedfactors.Finally,it are further connected to a backbone wired network. In this decideswhichtypeofaggregationissuitablefortheaggre- way, large areas can be covered by a low-cost infrastruc- gationi.e.,nodetonodeorendtoendaggregation.AADM ture. As an intermediary, a mesh access points (MAP) also alsodefinesroleofsourceanddestinationintheaggregation servesthenetwork,whichworksontheprincipleofaradio applied. frequency (RF) access point and transmits the packet to Therestofthispaperisorganizedinthefollowingorder: othermeshnetworks,ortothewiredbackbone.Ittakesthe Related work is elaborated with respect to types of aggre- packet form one mesh client network and forwards either gationapproachesinSect.2.Proposedframeworkhasbeen to some other mesh client network or some heterogeneous describedcontainingthedetailofaggregationalgorithmfor network using the wired backbone. MR chooses a suitable generatinghashtablesinSect.3,andthenresultsandanaly- gatewaytoroutethereceivedpackets.MGisanentitythat sisfollowinSect.4,andinthefinalsectionweconcludethe combines or integrates two networks like connecting the paper. wired network with the wireless mesh network for better support. Wireless mesh networks also inherit the characteristics 2 Relatedwork of ad hoc networks such as self-organized, self-configured, self-healing, scalability and reliability. Such kinds of net- Aggregation is a composition technique for building a new worksautomaticallyincorporateanewnodeintotheexisting objectfromoneormoreexistingobjectsthatsupportsome structure without requiring any adjustments by a network or all of the new object’s required interfaces [2]. There are administrator.Itsconfigurationsallowalocalnetworktorun twotypesofaggregationwithrespecttopacketsize.Oneis faster, because a local packet does not need to travel back static aggregation and the other is dynamic aggregation. In toacentralserver.Theycanprovidecost-effectivenetwork static aggregation, the maximum transmission unit (MTU) deployment under the diverse environments of PSDR and remains fixed, while in dynamic aggregation it may varies better coverage to both stationary and mobile users. They accordingtotheinputparameters.Thedynamicaggregation offer multiple paths from a source to destination through issimilartothefixedaggregationexceptthecharacteristics their on demand routing algorithms allowing each node to oflocallinkitusedtodetermineanappropriatepacketsizeto makeanintelligentdecisioni.e.,whichpathcaneffectively reducethechancesofnetworkpacketsbeingdropped.Such beusedtoforwardpacketsthroughthenetworkinorderto packetsizeiscalledtheaggregationthresholdandisnotmore improvetheoverallnetworkperformance. thanthatthevalueofMTU.Thedecisionwhentoaggregate WhileWMN’sdeploymentsforPSDRpurpose,efficient issubstantiallyinfluencedbytwoparametersi.e.,themaxi- utilization of the bandwidth is a critical issue especially mumqueuesizeandthetimedelay.Inthisregard,existing for audio, text, images or video traffic. Some types of data literaturecanbebroadlyclassifiedintothreecategoriesi.e., might be bearing small chunk of packets for control infor- frameaggregation,packetaggregationandthelinkaggrega- mation, though small but numerous packets give rise to tion. 123 JReliableIntellEnviron(2015)1:159–171 161 2.1 Packetaggregation mentofVoIPcapacitywiththemaintenanceofvoicequality. It helps reduce the MAC layer contention and significantly In packet aggregation a number of small packets are com- increasestheconcurrentVoIPflows.Itisahop-by-hopaggre- binedintolargerpackets.Inthisapproachdistributedaggre- gation scheme and inherently derives all the limitations of gators are used to collect smaller packets through various suchschemes.Theirmaincontributionistodesignthepacket network connections and assemble as a large packet. In aggregationmechanismwhichadaptsitselftonetworktraf- packet aggregation the sender adds an aggregation header, ficandminimizesdelayandoverheadwithoutrequiringany so that the receiver can de-aggregate the packets correctly, changes to current MAC layer implementations. So then, which cause this a more critical operation. There are two increasesVoIPperformanceandreducesMACdelay. typesofpacketaggregationi.e.,hop-by-hopaggregationand TheabovesolutionisfocusedonincreasingtheVoIPtraf- end-to-endaggregation.Inhop-by-hopscheme,thepackets ficinmulti-hopWMNs.Niculescuetal.[6]proposedseveral are aggregated and de-aggregated on each hop frequently methods to improve voice quality and used multiple inter- until packets become arrived at the final destination; while faces label based forwarding architecture along with path in end-to-end scheme packets are aggregated at source and diversityandaggregation.Theypresentexperimentalresults de-aggregatedatdestinationnodeonly. fromanIEEE802.11btestbed,optimizedforvoicedelivery. Bayeretal.[2]proposedahop-by-hoppacketaggregation Theyimplementedadistributedpacketaggregationstrategy mechanismfor802.11sWMNsandworkedonthefeasibil- andmadeagooduseofthenaturalwaitingtimeofMACfor ity of voice over internet protocol (VoIP) in a dual radio aggregationpurpose.Thesemethodsshowalittleimprove- meshenvironment.Theyintroducedanovelpacketaggrega- ment in the enhancement of simultaneous number of calls. tionschemetoavoidthesmallpacketoverheadthatreduces They focus on two important problems in supporting VoIP theMAClayerbusytimeandenhancesthenetworkperfor- overWMNsi.e.,increasingVoIPcapacityandmaintaining mance.ItaggregatespacketsatIPlayerandisenabledwith QoSunderinternalandexternalinterference.Theyevaluate differentnetworkconditionsandtrafficcharacteristics.This the performance of VoIP, over WMN and provide various approach does not increase delay unless it provides a good approachesforsystemoptimization. aggregationratio. Raghavendra et al. [7] proposed another method on an Kyungtae et al. [3] developed a distributed multi-hop IP-based adaptive packet concatenation for WMNs. In this aggregationalgorithmforVoMESHthatusesnaturalwaiting method authors used the adaptive technique i.e., to decide time in the interface queue of packets in a loaded network. during runtime whether to concatenate the packets or not. Theyusedacombinedapproachofheadercompressionand Thepacketsizeforaggregatedpacketsiscalculatedbasedon packetaggregationthatreducesVoIPprotocoloverheadand theroutequality,becauseagoodqualityroutecanpotentially introducessignallingoverhead.Theyproposedazero-length carrylargeraggregatedpackets. headercompressionalgorithmintegratedwithpacketaggre- In our findings, the aforementioned solutions have fol- gation, which does not need to depend on the signalling lowing shortcomings: [2,3] and [6] solutions are static and mechanismtorecoverthecontextdiscrepancybetweencom- areinefficienttouseifaggregationisnotrequired.The[2] pressor and de-compressor. The major challenges are to and [5] solutions ignore the bandwidth, energy consump- reduceVoIPprotocoloverheadand802.11MACoverhead, tion,congestion,routingandotherimportantfactorsinpacket whichdecreasesVoIPperformanceformeshnetwork.They aggregation. The [3,5] and [6] solutions cannot be applied investigate these problems in 802.11-based wireless mesh insinglehopWMNs.Since,[5]ignoresthesecurityofdata network and propose novel solutions to reduce overheads. andthenetworkatthesametime.The[5]and[6]emphasize Eachofthesemethodsproducesconsiderableimprovement on increasing capacity and QoS in VoIP, however, later in in operation of the mesh with respect to network capacity particular,didnotspecifyroutingmechanismofthepackets. andqualityofservice(QoS). The [7] ignores the influence of link quality on packet size Andreasetal.[4]proposedanadaptivehop-by-hopaggre- oronlyconsidertheend-to-endpathqualityduetotheuseof gationschemethatcomputesthetargetaggregationsizefor routingmetricssuchasWCETT.Thesemetricsreflectpath eachhopandisbasedonwirelesslinkcharacteristics.This characteristics,suitableforend-to-endaggregationandthus scheme behaves well relative to its counterpart schemes, achievesuboptimalperformance.Ifthereisabottlenecklink likestaticaggregationschemeforvariousperformancepara- suchasonecharacterizedthroughlowsignalquality,thiswill meters. They focused on the relation between link quality lead to a small packet size for end-to-end aggregation, and andpacketsizeforpacketaggregationinmulti-hopWMNs. thus the benefit of aggregating will be lost or negligible at Therefore, the overall aggregation along a path will not be all. constrainedbytheweakestlinkwhichleadstosignificantper- Furthermore,thesolutionsproposedforVoIPcannotsuc- formance improvement. Marcel et al. [5] proposed a novel cessfullybeimplementedinWMNduetotheabsenceofa packetaggregationmechanismwhichleadstotheenhance- centralcontroller.Secondly,themainpurposeofthesesolu- 123 162 JReliableIntellEnviron(2015)1:159–171 tionsistoincreasethenumberofcalls,improvevoicequality sion. The aggregation scheduler scheme by [17], presents and minimize overhead. Moreover, many other important adetailedsimulationstudy.Theproposedframeaggregation parameterssuchassecurity,congestion,androuting,reliabil- basedschedulerdynamicallychoosestheaggregatedframe ityandenergyconsumptionanddelayareignored.Therefore, size and technique, based on various relevant parameters. these solutions are not suitable for aggregation in WMNs. Theframeaggregationtechniqueshavetobeeffectiveeven Somesolutionsareparticularlysuggestedforwirelesssensor underunsaturatednetworkconditionswithlimitedthrough- networks(WSNs)[8–12].TheyarenotapplicableinWMNs put. This technique estimates the time deadline for frame becauseofmanifoldreasons.Firstly,thesesolutionsarepro- transmission.Inaddition,italsooutlinesamethodforselect- posed for WSNs, which cannot be implemented in WMNs ing the frame aggregation type. Lin et al. [18] conduct a because WSNs has a central controller while WMNs are thoroughstudyofAggregatedMACServiceDataUnit(A- essentiallylackinganycentraldevice.Secondly,thesolutions MSDU)andAggregate-MACProtocolDataunit(A-MPDU) proposedforWSNsmainlyfocusedonenergyconsumption. frameaggregationschemesinIEEE802.11n.Theyproposea Thirdly,alltheabovesolutionsarestaticandcannotdecide simpleoptimalframesizeadaptationalgorithmforA-MSDU onruntimewhetheraggregationisrequiredornotinagiven andA-MPDUundererror-pronechannels. scenario, which make these solutions inefficient especially Anotheradaptiveapproachisproposedby[19].Inwhich, forWMNs. authors used a cross-layer scheme that improves scalabil- ityofWMNsusingaggregationofMAClayerframes.The 2.2 Frameaggregation aggregationisperformedontopofMAClayer,allowingus to reduce the overhead caused by protocol headers and the Frameaggregationintegratesmanysmallerpacketstogether contentionmechanismregulatingtheIEEE802.11standard. to form a single large packet that could be sent in a single This approach relies on adaptive aggregation scheme and transmission and avoids the header of individual packets. leveragesthechannelprobingfunctionalitiesofmeshrouters. The frame aggregation scheme helps in improving the effi- Suchinformationisexploitedinordertocomputetheoptimal ciency for Media Access Control (MAC) protocol. Lee et saturationburstlength. al. [13] proposed a hop-by-hop frame aggregation scheme Some adaptive approaches are also proposed by [7] and for VoIP that reduces interference effect by combining the [19].Thesesolutionsareadaptiveinnaturebutjusttookvery inter-callaggregationandthepseudo-broadcastparameters. limitedparametersinconsideration.Alsothesesolutionsdid This scheme works effectively not just for the calls travel- notdecidewhichtypeofaggregationissuitableforthegiven ing same routed path but also for calls crossing each other scenario.Someauthorsin[20]and[21]haveproposedmech- insidethemulti-hopnetworks.Inthiswaythecallscouldbe anismstosolvetheperformanceanomalyproblembypacket entertainednotonlyforthesameroutingpathbutalsoforthe aggregationusingadynamictimeinterval,whichultimately pathscrossingwithotherroutingpathsinsidethemulti-hop dependsonthebusytimeofthewirelessmedium. networks. The shortcomings of above schemes, in our opinion, Lin et al. [14] proposes a dynamic frame aggregation are given as under: Firstly, a hop-by-hop frame aggrega- scheme called dynamic aggregation selection and schedul- tion (HHA) is proposed for VoIP in [13]. Secondly, the ingalgorithm.Thisalgorithmdynamicallyadoptsasuitable solution is focused on increasing in number of calls, band- schemeoutofdifferentaggregationmechanismsandisbased width utilization, routing mechanism, energy consumption on waiting for the next frame for the aggregation method. andotherimportantparametersareignored.Thirdly,hop-by- Kimetal.[15]proposedanothersolutionforframeaggrega- hopaggregationsolutionisnotefficientasaggregationand tion, which improves the throughput performance of IEEE de-aggregationisperformedateveryhop,whichutilizedthe 802.11WLAN.Theydevisedasimplemethodtoimplement networkresourcesrelativelyhigher. the frame aggregation into the real test-bed using off-the- Almostalloftheseframeaggregationschemesexcept[14] shelfIEEE802.11wirelessLANproductsviadevicedriver and [19] are misfit for WMNs due to fundamental differ- modifications.Thisschemeaggregatesmultipleframesintoa enceintheirnetworkarchitecture.Furthermore,allofthese singleMACframeinordertoreduceMAC/PHYoverheads schemesexcept[13]arestatic,whichincreasesinefficiency, of the IEEE 802.11 WLAN. Their current frame aggrega- consumes energy and bandwidth competitively more when tionschemeworkswhenframesaretransmittedtothesame aggregation is not required, and thus reduces the network receiver. life time due to higher energy consumption. At the same Gangulyetal.[16]proposedaframeaggregationscheme time these schemes cannot decide whether aggregation is forIEEE802.11WLANsinwhich,amodelisproposedfor requiredornot,whichmakesthesesolutionsinefficientfor calculatingsuccessfultransmissionprobabilityoftheframes a dynamic environment. In [14,15,15,17–19] schemes, the with certain length. Simulation results shown that network aggregatedframeswillbelostincaseofnetworkcongestion contention level have minor influence over the transmis- asthereisnoacknowledgmentindatalinklayerlikeofwire- 123 JReliableIntellEnviron(2015)1:159–171 163 lessnetworks.Inaddition,[15]reliesonacentralizedaccess Input point for being a centralized scheme and can be taken as a Values of parameters For Decision-1 flaw;likewise,[20]and[21]proposedsolutionforacomplete decentralizedinfrastructure. CalculateSum() NO While[17]scheme reduces theefficiency ofMAClayer >40 inanoisyenvironment.Atthesametimeitignoresenergy Yes Perform Aggregation consumption, congestion and reliability. In [14,15,15,17– Values of parameters 19]thedatalinklayerbasedsolutionisnotsecureenough. For Decision-2 CalculateAggregatedSum () In[19]theaggregationofframesincreasesthetransmission rate of data. Where, if the frame size is greater than MTU, <50 >50 data will ultimately be lost, as there is no option of frag- Hop-by-Hop End-to-End Aggregation Aggregation mentationindatalinklayerisavailable.Similarcaseholds for [14,15,15,17], and thus decrease the overall efficiency CalculateTrafficTypeSum () of the network. In [20] and [21], the authors used average Values of parameters For Decision-3 count,andsumforaggregationindynamicenvironment,but theysimultaneouslyignoresothersignificantconstraintslike energyconsumption,congestionandrouting. <50 >50 Real Time and Non Real Time Non Real Time 2.3 Link-basedaggregation CalculatePacketSize () Values of parameters Link aggregation aggregates many single links together to For Decision-4 formalinkaggregationgroup,sothatthewholelinkgroup would be treated as single link by any MAC client. In this <25 >50& <99 waythelogicallinksfromnodetonodemaybeestablished >25& <50 consisting more than one parallel instance of full duplex 2Packets 4Packets 3Packets point-to-point links. In this way the increased capacity of independentlinkaggregationfromaMAClayerislikelyto achieve. Output Okechetal.[22]proposedalink-basedaggregationframe- work and compares its performance with the fixed aggre- Fig. 1 FlowgraphforAADM gation and no-aggregation based schemes. They computes packet size by considering the link-based characteristics Figure1describestheflowgraphofproposedalgorithm and analyzed that VoIP performance can considerably be and the necessary steps that generate aggregated packet. In enhancedbyadjustingpacketsizewithlinkcharacteristics. which,itdescribedthatAADMfirstcalculatestheparame- terssumtofindthechancesofaggregationbeingperformed or not. In case the previous condition becomes true, it fur- 3 Proposedscheme thercalculatesthesumofparametersagaintoselectthetype of aggregation, and finally it calculates size of aggregated Inthispaper,weproposedaflexibleandefficientaggregation packets to be delivered to the destination. There are count- framework for WMNs, called adaptive aggregation-based lessfactorsthatcanbetakenintoaccountwithreferenceto decisionmodel(AADM).Inwhich,aggregationisbasedon aggregation. However, we emphasize a few of the primary various parameters i.e., congestion, delay, packet loss ratio factorslikecongestion,delay,packetlossratio,biterrorrate, (PLR), bit error rate (BER), buffer size and bandwidth. It buffer size and bandwidth. Here, we assume that all of the dynamically considers that aggregation is needed or not, if remaining factors are contributing average and acceptable neededandthenitwilldecidewhichtypeofaggregationis values, and are sufficient enough to make the aggregation suitable,i.e.,node-to-nodeorend-to-endaggregation.Later, feasible. Hence, keeping other entire factors constant with itdecideswhethersystemcouldaggregatebothrealtimeand respecttosourcenodesphysicalcontext,weshallfocuson non-real time data subject to the bandwidth support; other- the above mentioned primary factors. The impact of these wise,itprefersrealtimetrafficonly.Finally,AADMspecifies primaryfactorsissufficientenoughtocontributepositively MTUforaggregationbyinvestigatinglivestatisticsofdata towardstheaggregation. traffic.Moreover,itdefinestheroleofsourceanddestination In AADM, we developed an intelligent algorithm called intheaggregationmechanismapplied. aggregationfindingpossibilityalgorithm(AFPA)asdemon- 123 164 JReliableIntellEnviron(2015)1:159–171 strated bellow in Algorithm 1, which exclusively identify In turn, it gives the sum of parameters for compari- possibilityofaggregationbasedontheinformationprovided son with THRESH_LIMIT_AGG value. This threshold byanenvironment.Ifaggregationisnotfeasiblethenitrec- value will help in deciding the possibility of performing ommends transferring the packets as it is, otherwise AFPA aggregation. hastotakefewmoredecisions. Adaptive aggregation based decision model considers Inthisconnection,AFPAtellswhetheritshouldbehop- another threshold value namely THRESH_LIMIT_TYPE by-hopaggregationoritshouldbeend-to-endaggregation, that is set to 50 %, which suggests the type of aggrega- packets of which source and destination should be aggre- tion to be performed. The type of aggregation is selected gatedandsoon.Thesedecisionsmaketheaggregationmore fromhop-by-hopandend-to-endaggregation.Thehighsum efficient and results in delivery of more and more packets of input parameters value suggests the end-to-end type of byconsumingminimumnumberofresources.Thissolution aggregation;otherwisehop-by-hopaggregationisultimately alsohelpsinloadbalancingoverthenetworkandindirectly consideredtoexecute.Henceallthesethresholdvaluesallow playsitsroletoreducethenetworkcongestion. AADMwhilemakingsuchcriticaldecisions. ThepacketaggregationinAADMdecreasesphysicaland The varying degree of input values may lead in the MAC layer overhead and thereby reduces the transmission attainment of THRESH_LIMIT_AGG up to 40 % and time.Attimestheaggregationmightnotberequiredandwill progresstofurtherdecisionsrequired.Inonecaselessnet- bedecidedbyAADM.Theproposedalgorithmtakessomeof work congestion might be contributor in the achievement parametersastheinputandgetshelpstodecidewithathresh- of such threshold. In other case, if there is high conges- old value, say, THRESH_LIMIT_AGG, which is set as 40 tion, some other parameters like less average delay, higher inourscenario(assumed,andcanbeadjustedaccordingto buffersize,lesspacketdropratio,higherbandwidthorless the specification of network). If the sum of the parameters bit error rate, might be the major contributors in achieving isgreaterthantheTHRESH_LIMIT_AGG,theaggregation THRESH_LIMIT_AGG and vice-versa. The same thresh- will be performed. If the value of THRESH_LIMIT_AGG old is used to measure the possibility of real and non-real islessthan 40theoriginalpackets willbesentwithoutthe timetrafficaggregation.Therealtimetrafficwouldbeaggre- aggregationprocess.Ittakestheinputparameterslikecon- gated in every case while non-real time traffic depends on gestion,biterrorrate,delay,bandwidth,andbuffersizeand thereasonablevalueofthreshold.AADMconsidersanother packet loss ratio. If there is high congestion on the chan- thresholdcalledPACKET_SIZE_THRESHOLD(PST)i.e., nel,thealgorithmadaptivelydiscouragestheaggregationand PST_1,PST_2andPST_3.Thisthresholdestimatestherange vice-versa.Thehigherroundtriptime(RTT)ortheincreased ofmaximumpacketsize,inwhichthesumoftheparameters degreeofpacketdropduetotheshortqueuebufferindicates exists.Thelargersumwouldcertainlyleadtohigherthresh- the higher degree of congestion over the path. At the same old,andthusthehighernumberofpacketsperaggregation. time,ifthereismoreroutingdelaywhichisduetotheheavy Adaptive aggregation based decision model is funda- routingtables,noaggregationwillbeperformed. mentally based on fuzzy logic, as the real values of the Ifthereishighaveragedelayduetolessbufferorqueue parameters are passed as input arguments to the algorithm size,theaggregationisdisallowedbythealgorithm.Further- it transforms all these inputs into a suitable value for mak- more,ifthelinkisnotreliablewithrespecttobiterrorrate, ing decisions. Let us call it as the F_L_value for applying theaggregationwillnotbeperformed,asthelinkwithhigh fuzzylogic.Thesevaluescorrespondwiththeactualorreal biterrorratecannotberelieduponandmaynegativelyaffect parameters.TheseF_L_valuesarefurtherutilizedtocalcu- aboutintegrityofdata.Hence,thelinkquality(i.e.,inrelation latetheultimatethresholdvalue.Therearedifferentlevelsof toBER)isconsideredanothersignificantfactorindeciding F_L_values assigned to different parameters in every deci- theviabilityofaggregation. sion.Inmostcasestherelationshipisinverselyproportional Lastly, the number of the neighbours might affect such between the original input value and the assigned weight decision given that, the increased number of neighbours scale. performingaggregationwouldbeabottleneckfortheband- For instance, in the first case of decision if congestion width capacity. If the bandwidth is reasonably higher, the is very low i.e., 1 or 5 % we assign 0.15–0.20 points and aggregation probability would be relatively high. Hence, if congestion is very high i.e., 20–25 % then assign 0.01– thehighernumberofneighboursdoesnotnegativelyaffect 0.02pointsetc.Similarly,foreachparameterF_L_valuehas the system. The higher the packet loss ratio, the lesser beenassigned.Here,thetotalpointsare20outofwhichthe would be the chances of aggregation, which is one of F_L_valuesaregiven.Likewise,inpacketlossratioandbit the higher weight parameters that can aggressively affect error rate the total points are 25, in delay 10, in bandwidth the chances for aggregation. For all the decisions, there 10andbuffersizeisalso10.Thesemaximumpointsvaryfor is a reasonably higher weight scale for input parame- every decision and weight formula also changes for every ters, that is, varying values will be fed to the algorithm. decision. 123 JReliableIntellEnviron(2015)1:159–171 165 Algorithm 1AFPA: Aggregation finding possibility 1: AFPA (PLR, BER, D, B, BS, Cn){ 82: ------------ElseIf PLR > (5%) and < or = (10%) then p = 0.10 2: ------------If PLR < (1%) or = (1%) then p = 0.25 83: ------------ElseIf PLR > (10%) and < or = (15%) then p = 0.07 3: ------------ElseIf PLR > (1%) and < or = (2%) then p = 0.22 84: ------------ElseIf PLR > (15%) and < or = (20%) then p = 0.02 4: ------------ElseIf PLR > (2%) and < or = (5%) then p = 0.15 85: ------------ElseIf PLR > (20%) and < or = (25%) then p = 0.01 5: ------------ElseIf PLR > (5%) and < or = (10%) then p = 0.10 86: ------------If BER < (1%) or = (1%) then Lq = 0.20 6: ------------ElseIf PLR > (10%) and < or = (15%) then p = 0.07 87: ------------ElseIf BER > (1%) and < or = (2%) then Lq = 0.17 7: ------------ElseIf PLR > (15%) and < or = (20%) then p = 0.02 88: ------------ElseIf BER > (2%) and < or = (5%) then Lq = 0.12 8: ------------ElseIf PLR > (20%) and < or = (25%) then p = 0.01 89: ------------ElseIf BER > (5%) and < or = (10%) then Lq = 0.10 9: ------------If BER < (1%) or = (1%) then Lq = 0.25 90: ------------ElseIf BER > (10%) and < or = (15%) then Lq = 0.07 10: ------------ElseIf BER > (1%) and < or = (2%) then Lq = 0.22 91: ------------ElseIf BER > (15%) and < or = (20%) then Lq = 0.04 11: ------------ElseIf BER > (2%) and < or = (5%) then Lq = 0.15 92: ------------ElseIf BER > (20%) and < or = (25%) then Lq = 0.01 12: ------------ElseIf BER > (5%) and < or = (10%) then Lq = 0.07 93: ------------If D < 10ms or = 10ms then d = 0.40 13: ------------ElseIf BER > (10%) and < or = (20%) then Lq = 0.02 94: ------------ElseIf D > 10ms and < or = 100ms then d = 0.30 14: ------------ElseIf BER > (20%) and < or = (25%) then Lq = 0.01 95: ------------ElseIf D > 100ms and < or = 500ms then d = 0.15 15: ------------If D < 10ms or = 10ms then d = 0.1 96: ------------ElseIf D > 500ms and < or = 1s then d = 0.07 16: ------------ElseIf D > 10ms and < or = 100ms then d = 0.07 97: ------------ElseIf D > 1s and < or = 2s then d = 0.01 17: ------------ElseIf D > 100ms and < or = 500ms then d = 0.04 98: ------------If B < 512Kb or = 512Kb then b = 0.05 18: ------------ElseIf D > 500ms and < or = 1s then d = 0.02 99: ------------ElseIf B > 512Kb and < or = 1Mb then b = 0.05 19: ------------ElseIf D > 1s and < or = 2s then d = 0.01 100: ------------ElseIf B > 1Mb and < or = 2Mb then b = 0.04 20: ------------If B < 512Kb or = 512Kb then b = 0.1 101: ------------ElseIf B > 2Mb and < or = 4Mb then b = 0.03 21: ------------ElseIf B > 512Kb and < or = 1Mb then b = 0.09 102: ------------ElseIf B > 4Mb and < or = 8Mb then b = 0.02 22: ------------ElseIf B > 1Mb and < or = 2Mb then b = 0.07 103: ------------ElseIf B > 8Mb and < or = 16Mb then b = 0.01 23: ------------ElseIf B > 2Mb and < or = 4Mb then b = 0.04 104: ------------If BS < 100 or = 100 then q = 0.05 24: ------------ElseIf B > 4Mb and < or = 8Mb then b = 0.02 105: ------------ElseIf BS > 100 and < or = 500 then q = 0.04 25: ------------ElseIf B > 8Mb and < or = 16Mb then b = 0.01 106: ------------ElseIf BS > 500 and < or = 1000 then q = 0.02 26: ------------If BS < 100 or = 100 then q = 0.1 107: ------------ElseIf BS > 1000and < or = 5000 then q = 0.01 27: ------------ElseIf BS > 100 and < or = 500 then q = 0.07 108: ------------If Cn < (1%) or = (1%) then c = 0.10 28: ------------ElseIf BS > 500 and < or = 1000 then q = 0.04 109: ------------ElseIf Cn > (1%) and < or = (5%) then c = 0.07 29: ------------ElseIf BS > 1000 and < or = 5000 then q = 0.01 110: ------------ElseIf Cn > (5%) and < or = (10%) then c = 0.05 30: ------------If Cn < (1%) or = (1%) then c = 0.2 111: ------------ElseIf Cn > (10%) and <or =(15%) then c = 0.03 31: ------------ElseIf Cn > (1%) and < or = (5%) then c = 0.15 112: ------------ElseIf Cn > (15%) and < or =(20%) then c = 0.02 32: ------------ElseIf Cn > (5%) and < or = (10%) then c = 0.10 113: ------------ElseIf Cn > (20%) and < or =(25%) then c = 0.01 33: ------------ElseIf Cn > (10%) and < or =(15%) then c = 0.06 114: ----------------------------SRE = ∑( b, p, d, Lq, q, c) 34: ------------ElseIf Cn > (15%) and < or =(20%) then c = 0.02 115: IF SRE > THRESH_LIMIT_TYPE 35: ------------ElseIf Cn > (20%) and < or =(25%) then c = 0.01 116: --------Then “Aggregation can be performed on both Real Time and Non- 36: ----------------------------SAF = ∑(b, p, d, Lq, q, c) Real Time Traffic” 37: If SAF < THRESH_LIMIT_AGG 117: Else “Aggregation can only be performed on Real Time Traffic”}{ 38: ------------Then “Aggregation cannot be performed” 118: ------------If PLR < (1%) or = (1%) then p = 0.40 39: Else {{ 119: ------------ElseIf PLR > (1%) and < or = (2%) then p = 0.35 40: ------------If PLR < (1%) or = (1%) then p = 0.35 120: ------------ElseIf PLR > (2%) and < or = (5%) then p = 0.22 41: ------------ElseIf PLR > (1%) and < or = (2%) then p = 0.30 121: ------------ElseIf PLR > (5%) and < or = (10%) then p = 0.14 42: ------------ElseIf PLR > (2%) and < or = (5%) then p = 0.20 122: ------------ElseIf PLR > (10%) and < or = (15%) then p = 0.06 43: ------------ElseIf PLR > (5%) and < or = (10%) then p = 0.10 123: ------------ElseIf PLR > (15%) and < or = (20%) then p = 0.04 44: ------------ElseIf PLR > (10%) and < or = (15%) then p = 0.07 124: ------------ElseIf PLR > (20%) and < or = (25%) then p = 0.01 45: ------------ElseIf PLR > (15%) and < or = (20%) then p = 0.02 125: ------------If BER < (1%) or = (1%) then Lq = 0.40 46: ------------ElseIf PLR > (20%) and < or = (25%) then p = 0.01 126: ------------ElseIf BER > (1%) and < or = (2%) then Lq = 0.35 47: ------------If BER < (1%) or = (1%) then Lq = 0.35 127: ------------ElseIf BER > (2%) and < or = (5%) then Lq = 0.22 48: ------------ElseIf BER > (1%) and < or = (2%) then Lq = 0.30 128: ------------ElseIf BER > (5%) and< or = (10%) then Lq = 0.14 49: ------------ElseIf BER > (2%) and < or = (5%) then Lq = 0.20 129: ------------ElseIf BER > (10%) and < or = (15%) then Lq = 0.06 50: ------------ElseIf BER> (5%) and < or = (10%) then Lq = 0.10 130: ------------ElseIf BER > (15%) and < or = (20%) then Lq = 0.04 51: ------------ElseIf BER > (10%) and < or = (15%) then Lq = 0.07 131: ------------ElseIf BER > (20%) and < or = (25%) then Lq = 0.01 52: ------------ElseIf BER > (15%) and < or = (20%) then Lq = 0.04 132: ------------If D < 10ms or = 10ms then d = 0.05 53: ------------ElseIf BER > (20%) and < or = (25%) then Lq = 0.01 133: ------------ElseIf D > 10ms and < or = 100ms then d = 0.04 54: ------------If D < 10msor = 10ms then d = 0.05 134: ------------ElseIf D > 100ms and < or = 500ms then d = 0.03 55: ------------ElseIf D > 10ms and < or = 100ms then d = 0.04 135: ------------ElseIf D > 500ms and < or = 1s then d = 0.02 56: ------------ElseIf D > 100ms and < or = 500ms then d = 0.03 136: ------------ElseIf D > 1s and < or = 2s then d = 0.01 57: ------------ElseIf D > 500ms and < or = 1s then d = 0.02 137: ------------If B < 512Kb or = 512Kb then b = 0.05 58: ------------ElseIf D > 1s and < or = 2s then d = 0.01 138: ------------ElseIf B > 512Kb and < or = 1Mb then b = 0.05 59: ------------If B < 512Kb or = 512Kb then b = 0.05 139: ------------ElseIf B > 1Mb and < or = 2Mb then b = 0.04 60: ------------ElseIf B > 512Kb and < or = 1Mb then b = 0.05 140: ------------ElseIf B > 2Mb and < or = 4Mb then b = 0.03 61: ------------ElseIf B > 1Mb and < or = 2Mb then b = 0.04 141: ------------ElseIf B > 4Mb and < or = 8Mb then b = 0.02 62: ------------ElseIf B > 2Mb and < or = 4Mb then b = 0.03 142: ------------ElseIf B > 8Mb and < or = 16Mb then b = 0.01 63: ------------ElseIf B > 4Mb and < or = 8Mb then b = 0.02 143: ------------If BS < 100 or = 100 then q = 0.05 64: ------------ElseIf B > 8Mb and < or = 16Mb then b = 0.01 144: ------------ElseIf BS > 100 and < or = 500 then q = 0.04 65: ------------If BS < 100 or = 100 then q = 0.1 145: ------------ElseIf BS > 500 and < or = 1000 then q = 0.02 66: ------------ElseIf BS > 100 and < or = 500 then q = 0.07 146: ------------ElseIf BS > 1000 and < or = 5000 then q = 0.01 67: ------------ElseIf BS > 500 and < or = 1000 then q = 0.04 147: ------------If Cn < (1%) or = (1%) then c = 0.05 68: ------------ElseIf BS > 1000 and < or = 5000 then q = 0.01 148: ------------ElseIf Cn > (1%) and < or = (5%) then c = 0.05 69: ------------If Cn < (1%) or = (1%) then c = 0.10 149: ------------ElseIf Cn > (5%) and < or = (10%) then c = 0.04 70: ------------ElseIf Cn > (1%) and < or = (5%) then c = 0.07 150: ------------ElseIf Cn > (10%) and < or =(15%) then c = 0.03 71: ------------ElseIf Cn > (5%) and < or = (10%) then c = 0.05 151: ------------ElseIf Cn > (15%) and < or =(20%) then c = 0.02 72: ------------ElseIf Cn > (10%) and < or = (15%) then c = 0.03 152: ------------ElseIf Cn > (20%) and < or =(25%) then c = 0.01 73: ------------ElseIf Cn > (15%) and < or = (20%) then c = 0.029 153: ---------------------------SPS = ∑( b, p, d, Lq, q, c) 74: ------------ElseIf Cn > (20%) and < or = (25%) then c = 0.01 154: ------------If SPS < or = PST_1 Then 75: ----------------------------SHE = ∑( b, p, d, Lq, q, c) 155: --------------------“The sender can aggregate 2 numbers of packets” 76: IF SHE >THRESH_LIMIT_TYPE 156: ------------If SPS > PST_1 and < or = PST_2 Then 77: ------------Then “End-end-based approach is selected” 157: --------------------“The sender can aggregate 3 numbers of packets” 78: Else “Hop-based approach is selected”}{ 158: ------------If SPS > PST_2 Then 79: ------------If PLR < (1%) or = (1%) then p = 0.20 159: --------------------“The sender canaggregate 4 numbers of packets” 80: ------------ElseIf PLR > (1%) and < or = (2%) then p = 0.17 160: } 81: ------------ElseIf PLR > (2%) and < or = (5%) then p = 0.12 123 166 JReliableIntellEnviron(2015)1:159–171 The proposed algorithm has incorporated an end-to-end Table1 Notationsforinputparameters packet aggregation approach for voice-based packets and Symbol Description hop-by-hopaggregationapproachforothertypesofdatatraf- fic. The voice packets are required to get to the destination SAF Parameterssumforaggregationfeasibility onreal-timebasisandtheaggregationorde-aggregationof SHE Parameterssumforaggregationtypeselection packetstakestimeateachhop,whenhop-by-hopisselected. SRE Parameterssumofreal-timedataaggregation Hence, the later approach is discouraged for voice-based SPS Parameterssumforcalculationofpacketsize packets. However, in case of packets bearing diverse des- PST Packetsizethreshold tinations,thehop-by-hopisacrucialapproachandcannotbe overlooked.Incaseofvoicepacketshavingthesamesource Table2 Weightsdistributionschemesforattributes and destination, the end-to-end approach is adopted neces- sarily.Aftermakingsuchdecision,AFPAalgorithmmakes Parameter D-1 D-2 D-3 D-4 this distinction on packet-to-packet basis and selects a par- Bandwidth 0.10 0.05 0.05 0.05 ticularqueueaccordingly.Datapacketsaretreatedaccording Delay 0.10 0.05 0.40 0.05 torealtimeornon-realtimerequirement. PLR 0.25 0.35 0.20 0.40 The maximum size of the aggregated packet cannot be BER 0.25 0.35 0.20 0.40 morethanMTU,whichisdeterminedbyeverynetwork.If Buffersize 0.10 0.10 0.05 0.05 AADM selects to perform the aggregation, then, the level of aggregation is also adjusted according to the magnitude ofPACKET_SIZE_THRESHOLDasexplainedearlier.The MTUforwirednetworksis1500bytes,andforIEEE802.11 assignedtotheseattributesbecomesdifferentforeachdeci- is 2300 bytes. Hence, the aggregated packet size remains sionandarepresentedinTable2.AtfirsttheAFPAalgorithm below than MTU. It is therefore, in AADM the higher takes the input and adapts these values according the crite- thePACKET_SIZE_THRESHOLDthehigherwouldbethe rionspecified.Theseadaptedormodifiedparametervalues aggregationlevelandthesizeofaggregatedpacket.Besides, are added to find a sum to compare with stipulated thresh- these thresholds are adjustable parameters, and would vary olds.Inthisway,AFPAtakeusefuldecisionsanddetermine accordingtothespecificsoftheenvironment.Hence,itwould theactionsrequiredtoperformwithpacketsdynamically.We notbewisetoassumeasinglestaticthreshold,suitabletoany comeupwithmultipleactionsseriallyaswetraversedown particularsetting. byAFPAalgorithm. Fordecision-1,thepassingcriterionis40%,thatis,ifthe 3.1 SelectedattributesforAADM sumofparametersisabove0.40,theaggregationwouldbe performed; otherwise will not be considered. For decision- In aggregation, there are many attributes which play vital 2, the passing criterion is 50 %, given that, if the sum of roleindefiningtheefficiencyofameshnetwork.Therepre- the parameters is above than 0.50, then execute end-to-end sentative attributes are given as: congestion, delay, packet aggregation; otherwise, hop-by-hop aggregation would be loss ratio (PLR), bit error rate (BER), buffer size, and performed.Fordecision-3,thepassingcriterionisagainset bandwidth. In the sequel, we perform the following four to50%,inwhichifthesumofparametersisabovethat0.50, aggregation decisions on the basis of aforementioned six thenitwillfollowboththerealtimedataandnon-realtime attributes,i.e.,aggregation/noaggregation;hop-by-hop/end- data aggregation; otherwise only the real time data traffic to-end; real time/non-real time, and packet size. For each aggregationwouldbeperformed. decisionweallocateweightstoeachoftheabove-mentioned For decision-4, the MTU is specified by identifying the attributes.Sincetheroleofeachattributediffersineachof number of packets to be aggregated. The passing criterion the four decisions, their weights become changed accord- tendstobeprogressive,astheMTUsizeisgettinglargewith ingly. The list of input variables is given as: (1) bandwidth increasingsum.Ifthesumofparametersisupto0.25,then (b),(2)averagedelay(d),(3)packetlossratio(Lq),(4)bit onlytwopacketswouldbeaggregated;ifitisabove0.25and error rate (p), (5) buffer size_BS (q), (6) congestion (C). lessthan0.50thenthreepacketswouldbeaggregated,while Whereas, the constants defined for the proposed algorithm fourpacketswouldbeaggregated,ifthesumisabovethen are as follows: (1) THRESH_LIMIT_AGG = 40 %, (2) the0.50until0.99. THRESH_LIMIT_TYPE = 50 %, (3) PST_1 = 25 %, (4) PST_2=50%,(5)PST_3=99%. 3.2 Casestudy Table 1, further represents the notations and inputs for theAFPAalgorithm.Asdefinedthattherearefourdecisions Anetworkhas2Mbbandwidthwith10%congestionlevel, (D)totake,foreachdecisionmaximumweightsthatcanbe 10%packetlossratio(PLR),5%biterrorrate(BER),100 123 JReliableIntellEnviron(2015)1:159–171 167 msaveragedelay,and500buffersize.Theprospectforper- formingaggregationornotinanetworkisasfollows: 2Mb=0.07 10%=0.10 10%=0.10 5%=0.15 100ms=0.04 500=0.07 since, 0.53 (53 %), is more than 40 % so aggregation is allowed. And, the prospects for performing either hop-by- hoporend-to-endaggregation: 2Mb=0.07 10%=0.05 10%=0.10 5%=0.20 100ms=0.04 500=0.07 As0.50(50%)isequalto50%,thusend-to-endaggregation isallowed.Theprospectsforperformingeitherrealtimeor Fig. 2 Hashtableofeachattributefordecision-1 non-realtimeorboth: 2Mb=0.04 10%=0.05 10%=0.10 5%=0.12 100ms=0.30 500=0.04 and,0.65(65%)ismorethan50%,soaggregationcanbe appliedonrealtimeandonnon-realtimedata.Finally,the determinednumberofpacketstobeaggregated: 2Mb=0.04 10%=0.04 10%=0.14 5%=0.22 100ms=0.04 500=0.04 Inthiscasehere,4packetswillbeaggregated,asthecal- culatedsumisabovethat0.50i.e.,0.52(52%).Nevertheless, thefourhashtablesillustratedinFigs.2,3,4and5arebased ontheselectedsixparameters.However,themappingweight differs for each one. Our algorithm is based on the packet basedaggregation,ratherthanframeorlinkbasedaggrega- Fig. 3 Hashtableofeachattributefordecision-2 tion.Hence,theVoIPpacketisrelativelysmallerthanpackets of other types of data. The control information in the form ofheaderhastobesentwitheachpacket.Thesmallpacket traditionalnetworks.Hencetheschemeisorientedtowards meansmorecontrolinformationisbeingsentonthechannel finding the solution for VoIP streaming, which leads to the thatleadstoextraoverhead.TheVoIPdatasuffersthemost selectionofEnd-Endbasedapproach. 123 168 JReliableIntellEnviron(2015)1:159–171 Fig. 4 Hashtableofeachattributefordecision-3 Fig. 5 Hashtableofeachattributefordecision-4 Nowitisbindingontheingressnodeperformingaggre- gation, to explore the ultimate and alike destinations of maximum number of packets getting through it. However, inothercasetheHop-by-Hop approach wouldbethesolu- tion. Since single packet is not the right one candidate for aggregationi.e.,iftheingressnodefindsonlyasinglepacket ortwoforsomedestination,itshouldnotperformaggrega- tion in that case and would cost the network, if it is being Fig. 6 Arrowtopology performed.Thus,thisisoneofthecaseswhereaggregation isnotgoingtobeperformed,besidestheothercaseswehave seeninthealgorithm.Thealgorithmtakescareofdifferent parameters i.e., the number of packets accurately received, inputparameterswhichareassignedwithdifferentsuitable totaltimeconsumedandtotalnetworkload.Ifthethrough- fuzzylogicvalues.Finallyacomputedvalueisobtainedafter putisincreasing,thenthatwillindicatethatmorepacketsare summingupthefuzzylogicvaluesandis,thencomparedwith accuratelyreceived,lesstimeisconsumedandthereisless athresholdvariabletofindthefeasibilityofaggregation. networkload. Weperformedsimulationsofthedesignanddevelopment withintheframeworkofanopenscalablewirelessmeshnet- 4 Resultsandanalysis work,byusinganarrowbasedmixedtopologyasshownin Fig.6.Welinkednode1tonode6indirectly,withthecom- This section describes simulation setup and discussion of bination of wired and wireless. The node 1 is linked with theresultsobserved.Inthesequel,tomakecomparisonwith nodes2and3usinggigabitEthernet(1000Mb/s).Thenode proposed scheme, out of the existing schemes as discussed 3isconnectedwithnode4.Theformerworksasagateway, in Sect. 2, we have selected one of the static scheme pro- whilethelaterperformsasmeshrelaytowardsnodes5and posedby[17],duetothecommonapplicationenvironment. 6,bothinclusive. WeusedOMNeT++simulatortomodelandverifyAADM. TheMTUistakenas1500bytes.Inparticular,theMAC Beforewediscussthesimulationsetup,wedefinethrough- protocolIEEE802.11risselected,whichisdesignedwitha put in our scenario first, which is the combination of three specialfocusonVoIP.Thesimulationwasaverageoftenrun 123

Description:
Flexible and efficient aggregation framework for antifragile wireless mesh networks. Muhammad Shahzad antifragile and resilience application framework that can also provide on-demand computing . The packet size for aggregated packets is calculated based on the route quality, because a good
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.