894 Ixp Research PDF

8/3/2019 894 Ixp Research PDF

http://slidepdf.com/reader/full/894-ixp-research-pdf 1/48

InternetExchangePoints AcloserlookatthedifferencesbetweencontinentalEuropeandtherest

oftheworld.

Stephanie Silvius

Amsterdam, January 2011



2

TableofContents

Abstract.................................................................................................................................4

1 Euro-IX............................................................................................................................5

2 Theresearch.................................................................................................................63 Transit,peeringandIXPs.........................................................................................73.1 HowtheInternetworks.................. ................. .................. .................. .................. ........73.2 AutonomousSystems(AS)................ ................. .................. .................. .................. .....73.3 Tierhierarchy.................. .................. ................. .................. .................. .................. ........93.4 Transitvs.peering........................................................................................................10 3.5 WhatisanIXP?..............................................................................................................10

3.5.1 ThepurposeofanIXP..........................................................................................................113.5.2 ConnectingtoanIXP............................................................................................................12

4 AbriefhistoryofIXPs.............................................................................................134.1 Europe..............................................................................................................................13

4.1.1 WesternEurope....................................................... ............................................................. ..134.1.2 NorthernEurope....................................................................................................................144.1.3 CentralEurope........................................................................................................................144.1.4 SouthernEurope....................................................................................................................154.1.5 EasternEurope.......................................................................................................................15

4.2 NorthAmerica................................................................................................................16 4.3 Asia....................................................................................................................................17

4.3.1 China............................................................................................................................................174.3.2 Japan............................................................................................................................................194.3.3 India.............................................................................................................................................194.3.4 Nepal............................................................................................................................................20

4.4 LatinAmerica.................................................................................................................20 4.4.1 Brazil...........................................................................................................................................204.4.2 Argentina...................................................................................................................................21

4.5 Africa.................................................................................................................................22

5 IXPmodels..................................................................................................................245.1 Commercialvs.non-commercial.............................................................................24

5.1.1 Neutrality..................................................................................................................................245.1.2 Ownershipandmanagement........................................................... .................................245.1.3 Fees..............................................................................................................................................255.1.4 Scopeofactivities..................................................................................................................255.1.5 Peeringagreementsandpolicies....................................................................................26

5.2 Europe..............................................................................................................................29 5.3 NorthAmerica................................................................................................................29 5.4 Asia....................................................................................................................................30 5.5 LatinAmerica.................................................................................................................30 5.6 Africa.................................................................................................................................31

6 Currentdevelopmentsandtrends.....................................................................326.1 IXPparticipantsPointofView..................................................................................32 6.2 Trends...............................................................................................................................33

6.2.1 Japan............................................................................................................................................34

7 Differences.................................................................................................................357.1 Connectedparties.........................................................................................................35 7.2 Influenceincumbents..................................................................................................36 7.3 NorthAmericanvs.EuropeanIXPs.........................................................................36



3

7.4 DifferenceinASNsinEuropevs.US........................................................................37

8 Conclusion..................................................................................................................39

Glossaryofterms............................................................................................................40

Acknowledgements........................................................................................................41

Bibliography.....................................................................................................................42



4

Abstract

ThisdocumentdescribestheresearchthathasbeendonebetweenAugust2010

andJanuary2011fortheEuropeanInternetExchangePointassociation.Euro-IXgathersresourcesabouttheEuropeanIXPscenetobeabletomakefuture

predictionsandimprovetheIXPcommunity.TheresearchstartedatlookingintothedifferencesbetweenIXPsintheUSandEurope,howeverthissoondeveloped

intolookingatthedifferencesbetweenIXPsaroundtheworld,asitwasthought

thattheinitialresearchwouldbeeitherverylimitedorhadtobetoothoroughfortheavailabletime.Bylookingattheentireworldabetterunderstandingof

theglobalIXPscenewouldbegainedwhichwouldalsobeofuseforEuro-IX,

sincetheyhavemembersfromoutsideofEurope.

ProperknowledgewasneededonthepurposeandworkingsofIXPsandthiswasgainedrapidly.AfterthisabriefhistoryonIXPsinallcontinentswasmadeto

understandtheirhistoryandcurrentsituationalittle.AfterthisthedifferencesbetweenIXPsweremappedandlocatedineachcontinent.Asafollowupcurrenttrendsanddevelopmentswerelookedinto,whichledtobeingabletomake

somefuturepredictions.

Afterdoingtheresearch,itcanbesaidthatit’sverydifficulttocomparethe

differentIXPstoeachother,sincetheyaresimplytoodifferent.OverthenextcoupleofyearswecanseemoreIXPsbeingestablishedinLatinAmerica,Africa

and(partsof)Asia,wherecurrentlythereareveryfewIXPspresent.TheseIXPs

willprobablybeofanon-profitnature,whileinthefuturemorecommercialIXPsmightalsoshowup.InNorthAmericacommercialIXPsaredominant,whilein

Europetherearemorenon-profitIXPsbutthenumberofcommercialIXPsisincreasingandislikelytocontinuetoexpandoverthenextcoupleofyears.

InNorthAmericatheconnectionbetweennetworksaremainlyprivateinterconnections,whileinEuropepublicpeeringismorecommon.European

IXPsarelikelytooffermoreservicestowardstheirparticipants,whichmight

alsobethecaseforthenowless-developedregions.MorepeeringrelationshipswillbeestablishedwiththeAsianIXPscene,thoughthiswillprobablybeaslow

developmentduetothehightransitpricesinAsia.

Itcanbesaidwithsomecertainty,thatalargeincreaseinexchangedbandwidthwillbeseenoverthenextcoupleofyears,whichwillleadtoaneedofIXPplatformupgradingtokeepupwiththisdemand.



5

1 Euro-IX

InMay2001EuropeanInternetExchangePoints(IXP)inEuropedecidedthere

wasneedtocombinetheirresourcestostrengthen,developandimprovetheIXPcommunity.ThisledtotheformationoftheEuropeanInternetExchange

Association(Euro-IX).ThisdevelopmentandimprovementoftheIXPcommunityisachievedbythecoordinationoftechnicalstandards,thedevelopmentof

commonprocedures,sharingideas,knowledgeandsharingpublicstatisticsand

informationbetweenIXPs.TohelpaccomplishthisEuro-IXorganisesbiannualtwo-daylongmeetingsthatallEuro-IXmemberscanattend.Theyalsoprovide

mailinglistsandon-lineresources.

Euro-IXcurrentlyhas56affiliatedIXPsfrom37countriesaroundtheworldand

islocatedinAmsterdam,TheNetherlands.

ThereisinterestinhavingaclearerunderstandingofthedifferencesbetweenthecontinentsasthiscanhelpgiveabetteroverviewofthecurrentsituationofIXPsanddevelopmentsaroundtheworld.Oneofthequestionsthathavebeen

raisedforsometimenow,iswhythereisalargedifferencebetweentheamount

ofuniqueASNsatIXPsinEuropecomparedtotheamountconnectedtoNorthAmericanIXPs,whilethetotalnumberofadvertisedASNsonbothcontinentsis

quitesimilar.Thisisoneofthequestionsthathavebeenexaminedinthisresearchandanattempttoanswerthisquestionwillbemade.



6

2 Theresearch

TheEuropeanInternetExchangeAssociation(Euro-IX)hasbeengatheringdata

aboutIXPs,thepartiesconnectedtothemandinfrastructuretoexchangethetrafficforalmost10yearsnow.Thishasallowedthemtoproducereportsonthe

currentdevelopmentsandtrends,whichtheynowwanttousetoproducemoreinsightsintotrendsthathavebeenoccurringandworkonfuturepredictions.

Euro-IXisinterestinginusagetrendsthatareoccurringinotherregionsthan

EuropeandhowthesecomparetoEurope.WhattheIXPslooklikeintheUS,iftherearedifferencesinthewaythatInternettrafficishandled,ifAsianISPsare

aslikelytomakeuseofIXPsasthoseinEuropeandiffuturepredictionsin

differentregionslooksimilar.

Euro-IXhopestogainabetterpictureofthedifferencesbetweentheEuropeanandUSpeeringecosystems,whathasandstilldoesinfluencetheamountof

organizationsthatarepresentatpublicpeeringpointsonbothcontinents,what

makesEuropeanIXPs,relativelyspeakingmorepopularthantheircounterpartsintheUS?Withabetterunderstandingofthesepeeringecosystemswehopeto

betterunderstandandpredictthefutureoftheglobalpeeringscene.Ithastobe

deliveredinadocumentdescribingtheresultsandoutcomes.Euro-IXwishestohavebetterinsightsintrendsoccurringaroundIXPs,notonlyinEuropebutalso

intheUS.Tobeabletodothis,betterinsightsareneededofthepastandcurrentsituationonthiscontinent,whatdifferencesthereareandhowtheycompareto

eachother.

ItisassumedthatthemaindifferenceinIXPecosystemsbetweenEuropeandthe

USisthatintheUSmajorpeeringpointsareinfactrunbycolocationoperatorswiththegoalofmakingaprofitviaofferingcollocationspaceratherthansimply

servingthecommunityandthishasadirectaffectontheamountofASNsseenin

theUS.



7

3 Transit,peeringandIXPs

Internetisavailableallovertheworld,buthowdoyoureachawebsitethatis

locatedonadifferentcontinent?Wheredoesthetrafficgowhenyoumakeaninformationrequestandhowdoesitreachitsdestination?Allofthisisdoneby

InternetServiceProviders(ISPs)whoeitherpeerorbuytransitandisassistedbyInternetExchangePoints(IXPs).Thischapterexplainsthebasicsofhow

Internettrafficistransportedovertheworld.

3.1 HowtheInternetworks

Thedefinitionof‘TheInternet’fromDictionary.com[1]: “avastcomputernetworklinkingsmallercomputernetworksworldwide(usuallyprec.bythe).TheInternetincludes

commercial,educational,governmental,andothernetworks,allofwhichusethesameset

ofcommunicationsprotocols.”

WhenyouaresittingbehindyourcomputerandareconnectedtotheInternet,

youcantypeinaURLinyourwebbrowsertorequestawebpage,hitenterandalmostinstantlytherequestedwebpageappearsonyourscreen!Orwhenyou

messageafriendwithachatclient,yourmessageappearsalmostimmediatelyonhisscreenandhecanevenmessageyouback.What’shappeningbehindthe

scenesiswhatwecallthemagicoftheInternet .

Themomentyouhitenter,yourrequestorthemessageyouwanttosend,istransportedtoyourISP.ThisisthecompanythatyoubuyyourInternetaccess

from.AsacustomerofanISP,youarepartoftheISPs’network.Whenyourequestawebpage,itwillbetransportedtoyourISP,fromwhichitwillbe

transportedtoothernetworks(‘TheInternet’).Beforeyourrequestreachesits

destinationitprobablywentthroughalotofdifferentnetworks(dependingonwhereyourrequestisdestined).Allofthesenetworksthatyourrequestmay

travelthrougharecalledAutonomousSystems.OneAutonomousSystemmaybe

madeupofsmallernetworks,butthisisnotalwaysthecase.AlloftheseAutonomousSystemsconnectingtogethermaketheInternet,aswe

knowittoday.

3.2 AutonomousSystems(AS)

Throughouttheworldtherearenetworksthatareusedtodistributetrafficto

andfromothernetworks.ThesenetworksarecalledAutonomousSystems(AS)[3].EveryAShasauniqueAutonomousSystemNumber(ASN),whichisusedto



8

identifythenetwork.IngeneralISPsandcontentprovidershavetheirown

uniqueASN,whichisusedtoidentifytheirnetworktosetupBGP[2]sessions.Thismakesitpossibleforothernetworkstoconnecttothemandexchange

traffic.HavinganASNisalsoarequirementofconnectingtoanInternetExchangePoint(IXP).

Figure2isanexampleofhowdifferentASesareconnectedtooneIXPwithpeeringrelationshipstoanotherAS.Therulesofwhotopeerwith,usingtheIXP,

reallydependsontheIXPspeeringrules.Somewantalltheirparticipantstopeer

witheverybody,whileothersleavetheoptionofwhotopeerwithuptotheparticipantsthemselves.Settingupa(peering)relationshipwithanotherISP(or

contentprovider)isusuallyenteredintotodecreasethecostsforinterconnection.Furthermore,thenumberofAShopsthatarerequiredforthe

traffictotraverseandreachitsdestinationcanbereduced.IngeneralIXPsdon’t

interferewiththerelationshipsthatarecreatedbetweentheirparticipants.



9

3.3 Tierhierarchy

WhileingeneralInternetServiceProviders(ISPs)offerInternetaccesstotheir

customers(home-usersorbusinesses),thereisadifferenceinthe‘status’ofanISP.The‘status’ofanISPiscalledaTier.Whilethereisnoofficialdefinitionof

whatexactlyaTier1ISPactuallyis,itisgenerallythoughtofasanISPthatcan

reachanyothernetworkontheInternetwithoutthepurchaseoftransitoranyotherformofpayingsettlements.ThismeansTier1ISPsonlypeerwithother

Tier1ISPsandrequireallotherISPstopaytransittothemfortransportingtheirtrafficandmakethemaccessibletotherestoftheInternet.

Tier2ISPsareusuallyregionalInternetServiceProviders,theytendtopeermostoftheirtrafficwithothernetworks.Whentrafficcannotreachits

destinationviapeering,theISPwillhavetopurchasetransitfromanUpstream

ServiceProvider(Tier1ISP)togainconnectivitytoothernetworks.

ATier3ISPisusuallyknownasalocalInternetServiceProviderthatsolelypurchasetransitfromothernetworkstomakeconnectivityfortheircustomers

available.TheyeitherbuytransitfromtheirUpstreamServiceProvider,which

canbeeitheraTier1oraTier2ISP,ortheypeerwithothernetworks.

Figure3showsaTier1toTier3hierarchy.



10

3.4 Transitvs.peering

TrafficcanbetransportedovertheInternetintwoways:overtransitlinesorvia

peering.Transitisusedbetweentwoparties,usuallyfromanISPtoanUpstreamServiceProvider(usuallyTier1orTier2)totransportthetrafficviaapaid

agreement.TheISPundercontracthastopayfortrafficthatgoesoverthelinein

bothdirections,sobothincomingaswellasoutgoingtraffic.TheISPpaystheUpstreamServiceProvidertotransportthetrafficfurthertoitsfinaldestination

andtohaveabetterconnectivitytotheInternet.

Peeringisknownasawaytotransporttrafficbetweentwonetworksfor‘free’,

howeverinsomecases‘paid’peeringisalsopractised.Whentwopartiespeerandtheyhaveanequivalentamountoftrafficexchanged,thereisusuallyno

paymentinvolved.Whenthetrafficratiosaren’talike‘paid’peeringcanbean

alternative,thenetworkreceivingmostincomingtrafficwillgenerallyreceivepayment.Apartfromthecostsforpeering(eitherfreeorpaid),thenetwork

infrastructurebetweenthetwopartiesneedstobemaintainedaswell,usuallybothpartiesareresponsibleformaintainingtheinfrastructure.Apeering

relationshipisverycommonandusefulbetweentwopartiesthatsendalotof

traffictoeachother,iftheyweretopaytransititcouldpotentiallycostalotmoremoneyandtheconnectedpartythathastopaydoesn’thavethatmuchinfluence

onforexample,lineupgradesorswitches(whichtheyrequirewhentheygrowrapidly).

Therearetwodifferenttypesofpeering:publicandprivatepeering.Whenthree

ormorepartiesdecidetopeerpubliclytheywillconnecttoasinglepoint(switch).ThispublicpointofinterconnectionisalsocalledanInternetExchange

Point.Privatepeeringhappenswhentwonetworksdecidetointerconnect

directlywitheachother.Thisisusuallydoneatdatacenters(alsoknownascarrierhotels),whereadirectconnectioncanbeprovisionedbetweentwo

participants,bothlocatedinthesamebuilding.Thisrequiresaphysicalcabletobelaidbetweenthetwoports/switches/networkstomakeprivatepeering

possible.Mostofthetrafficthatgoesbetweenthelargestnetworksis

accomplishedviaprivatepeering.

3.5 WhatisanIXP?

IXPisanacronymforInterneteXchangePoint,whichisaphysicalpointwhereInternettrafficcanbeexchanged.Thiscanbesomethingassimpleasaswitchin

abasement,toalargenetworkspanningmultipledatacenters,withpower

backupandcoolingsystems.AnIXPprovidesthephysicalinfrastructurewhereISPs,universities,incumbents(Telco’s),contentprovidersandmanymore

entitiescanconnecttoeachother,bysettingupa(peering)relationshiptoexchangetraffic.

Thewordweusenowadaystodescribethephysicalpointofinterconnectionhaschangedovertheyears.Intheverybeginningwhenthefirstscientificnetworks

werebeingcreated,theycalledtheseinterconnectionpointsNetworkAccessPoint(NAP).LaterthischangedtobeingaMetropolitanAreaNetwork(MAE)

whenIXPsbecamecommercialandinthepresentdaysweknowit,particularlyinEurope,asanInternetExchangePoint(IXP).However,theyallmeanthesame



11

thing:aphysicalpointwheredifferentnetworkscanexchangetrafficwitheach

other.

InFigure4youcanseedifferentpartiesconnectedtoaswitchatanIXP.WhentheIXPisnon-profitthepartiesconnectedtotheIXParegenerallyreferredtoas

members.However,whentheIXPiscommercial,itsparticipantsarereferredtoascustomers.InthispaperwerefertothepartiesconnectedtotheIXPasparticipants,irrespectiveofwhethertheIXPisnon-profitorcommercial.

3.5.1 ThepurposeofanIXP

TheideaofanIXPwasstartedasawayofkeepinglocalInternettrafficlocal.Thismeans,notusinginternational(transit)linestoreachadestinationinthesame

regionaswheretherequestoriginallycamefrom.IXPswereintroducedsothatallpartieswillingtoparticipatecouldconnectatone(local)point,exchange

traffic,savemoneyanddecreaseRoundTripTimes(RTT).UsinganIXPalsoresultedinbetternetworkperformanceandQoS(QualityofService).

InternationallinesarealwaysavailabletouseincaseanIXPhappenstofail,so

theend-userwon’tnoticemuchintermsofperformanceshouldthisoccur.WhencustomersofanISPenjoytheirInternetexperience,theyaregenerallywillingto

paymorefortheirconnectivityandspeed.Moneypaidbycustomerscanbe

investedintheinfrastructureandservicestothecustomersandnotsimplyininternationaltransitcosts.Thisisgoodforthelocaleconomy.Withfewercosts





13

4 AbriefhistoryofIXPsInthissectionabriefintroductionwillbegiventocoverthehistoryandthe

presentdaysituationofIXPSinregionsaroundtheworld.

4.1 Europe

ManyofthefirstInternetconnectionsinEuropewereinitiatedbyacademic-andresearch-networks[4]orbytelecomoperators.Togethertheygatheredtocreate

onelocalpointofinterconnection,whichresultedinIXPs.IntheearlydaysofInternetExchangesinEurope,therewaslesscompetitionbetweenISPs,because

thebiggesttelecommunicationsoperator(Telco’s)keptthecostsoftransport,bothfortransitasfortransatlantic,veryhigh.Mostofthetrafficaroundthistime

wasdestinedforAmerica,becauseallcontentwasavailablethere.ThelocalISPs

wereforcedtoworktogethertoreducethecostsfortransportingtraffictoNorthAmerica.AfterafewyearsmorecontentbecameavailableinEurope,because

therewasmorelocalinterestandthisstartedtoplayapartinthedevelopment.

Theavailabilityoflocalcontentchangedthetrafficflow.MoretrafficstayedwithintheEuropeanbordersandlesstrafficwentoverseas,thismostlikely

playedanimportantroleinformingtheinterconnectionlandscape.TheveryfirstEuropeanIXPswereestablishedin1993.Thiswasaroundthetimewhen

Internetaccessbecamemorereadilyavailabletothehomeusersandmore

trafficstartedbeingexchanged.Moretraffichadalocaldestination,whichresultedintheneedforalocalexchange.Peeringhelpedtokeeptheend-user’s

costslow,thetransitandtransportcostswerehighwhileatthesametimethe

ISPsstayedcompetitivetowardsNorthAmericanbasedproviders.Today,theInternetpenetration[5]inEuropeis58.4%,andmakesup24.2%oftheInternet

usersworldwide.

4.1.1 WesternEurope

InthisdocumentwedefineWesternEuropeas:Belgium,France,theNetherlands,

LuxembourgandtheUnitedKingdom.

TheveryfirstInternetExchangesinWesternEuropewereestablishedin1994.

TheseIXPswerenamedLINX(LondonInternetExchange)andAMS-IX(AmsterdamInternetExchange).Today,theybothhavemorethan300

connectedmembersfromdiverselocationsaroundtheworld,withtrafficpeaksofover1.1TbpsatAMS-IX[6]andmorethan700GbpsatLINX[7].Thereare

approximately30IXPsspreadacrossWesternEurope.WhenlookingattheIXPs

inFrance,youcanseethatalmostalltheirIXPsarelocatedinandaroundParis,withtherecentcreationofFrance-IXsupportedbyAkamai,JaguarNetwork,Neo

TelecomsandInterxioninthehopeofgainingbetterconnectivityaroundFrance.

TheInternetpenetration[5]inWesternEuropeis2ndhighestinEurope,UnitedKingdomhas82.5%,theNetherlandswith88.6%,Belgium77.8%,Luxembourg

85.3%andFrance68.9%.





15

penetrations[4]isbetween58.4%and75.3%(Germanynotincluded).It’s

interestingtoseethatthecountrywiththelowestInternetpenetration(Polandwithjust58.4%)hasoneofthebiggesttrafficexchangingIXPsinCentralEurope,

whileSlovakia;withtheirIXPexchanginglesstraffic(about40Gbps)hasanInternetpenetrationof74.3%.

4.1.4 SouthernEurope

InthisdocumentwedefineSouthernEuropeas:Albania,BosniaandHerzegovina,Bulgaria,Croatia,Greece,Macedonian,Italy,Portugal,SerbiaandSpain.

MostoftheSouthernEuropeanIXPswereestablishedaroundthelate1990’s.In1997ESPANIX(Spain’sInternetNeutralExchangePoint)wasestablishedin

MadridconnectingthemostimportantSpanishtelecomoperatorsandare

currentlyoneofthelargestIXPsintheSouthernEuropeanregion.OthercountrieswithnotableamountsoftrafficattheirIXPsareBulgaria,Italyand

Greece.TheInternetpenetration[4]inSouthernEuropeissignificantlylower

comparedtothatoftheNorthernandWesternEuropeanregions,hereitliesbetween40%and65%,whileBosniaHerzegovinaatjust31.2%.

4.1.5 EasternEurope

InthisdocumentwedefineEasternEuropeas:Belarus,Estonia,Latvia,Lithuania,

Moldova,Romania,RussiaandUkraine.

MosttheIXPsinEasternEuropewereestablishedaround2000and2001.When

lookingatallthecountriesinthisregion,youcanseethatthetrafficexchangedinUkraineandRussiaaretremendouslyhighcomparedtotheothercountries,

wheremostofthetimethetrafficisbetween1Gbpsand10Gbps.TheUA-IX

(UkrainianInternetExchange)exchangesabout220GbpsatitspeakandtheMSK-IX(MoscowInternetExchange)haspeaktrafficofaround400Gbps.It’s

interestingtoseethatquiteafewEasternEuropeantelecomoperatorsandbusinessesconnecttoDE-CIXinGermanytogetmoreroutestotherestofthe

Internet.ThisisprobablyduetoGermany’sgeographicalproximitytothe

EasternEuropeanregion.

TheInternetpenetration[4]inEasternEuropeiscomparativelylowertothatof

otherEuropeanregions,howeverit’sgrowingrapidly,thusthepenetrationwillpotentiallybeevenhigherinthenextcoupleofyears.Estonia,Latviaand

LithuaniahavethehighestInternetpenetrationwith75.1%,67.8%and59.3%respectively.InBelarusthisfigureis46.2%,Moldavia30.0%andRomania

35.5%.It’sinterestingtonotethattheInternetpenetrationissomewhatlowerin

Russiawithjust42.8%andUkrainewith33.7%,whiletheamounttrafficattheirIXPsissubstantiallyhigherthanothercountriesinnotonlytheregionbutthe

wholeofEurope.



16

4.2 NorthAmerica

TheNorthAmericanhistoryofInternetissometimesreferredtoas‘the’

beginningoftheInternet.PeopleoftenseemtorefertothehistoryofARPANETandNSFNET[9].ThereasonforthisprobablybeingthatNorthAmericaisseen

asoneofthefirst

countrieswithInternetaccess.Thehistoryof

NorthAmericagoessomethinglikethis:

Inthelate1960stheUS

DepartmentofDefense’sAdvancedResearch

ProjectAgency(ARPA)

wasunsatisfiedhowthemanyresearch

institutionsanduniversitiesusedlocated

aroundthecountry,were

unabletoeasilyexchangeresultsofresearchwith

eachother.Thiswasduetothefactthatall

institutionsand

universitiesusedtheirowncustomizedcomputers,who Figure5.2showstheNorthAmericanstates

fittherequirementsfortheirownthathaveoneormultipleIXPs.Theredstate

research.Tomaketheexchange (NewYork)hasbeencalculatedtohavemoreofresultseasier,ARPAwanted than6IXPs.Thegreenstate(California)was

anetworkthatcouldeasily countedtohaveabove10IXPs.connectallthedifferentcomputer

systemstoeachother,sothatexchanginginformationwouldbealotfasterand

easier.ThisnetworkwaseventuallydevelopedandnamedARPANET.

Inthelate1980’sanewnetworkreplacedtheARPANET,becauseitcouldn’t

sufficetheneedsanddemandsoftheresearchersandmanyuniversitiesanymore.ThenewnetworkwassponsoredbytheNationalScienceFoundation

(NSF)andcomprisedoffiveNorthAmerica.TogethertheyformedtheNSFNETBackbone,basicallyformingaspinefortheInternet.In1989FederalInternet

Exchanges(FIX)werebuiltontheEastandWestCoast(FIXEastandFIXWest).

SoonafterthisthefirstCommercialInternetExchange(CIX)wasbuiltontheWestCoastofNorthAmerica,duetothefactthatthepeopleinchargeofthe

FIXesweretroubledaboutconnectingcommercialnetworkstotheNSFNET,theywantedtokeepthetwooftheseseparate.In1992MetropolitanFiberSystems

(MFS)builtthefirstMetropolitanArea(MAE)networktheWashingtonDCarea.

Thisquicklygrewintoaplacewheremanydifferent(commercial)networkswereabletointerconnectwitheachother.

Beforethe1990stheInternetwasmainlyusedforresearchpurposes.However,

thischangedwhenemailpenetratedintothegeneralcommunity.MoreandmorebusinessesandorganizationssawthebenefitoftheInternetanddecidedto



17

connecttothenetwork.ThisbecameaburdenfortheNSFNET,asthereweretoo

manypeoplewhilenotenoughcapacitytohandleallthistraffic.Toensureconnectivitytoallconnectedcompaniesandorganizationsthe

NationalScienceFoundationawardedcontractstofourNetworkAccessPoints(NAP).DifferentcompaniesspreadacrossNorthAmericaranalltheseNAPs.The

companiesthatreceivedthesecontractswere:ThePacificBellNAPinSanJose(California),theAmeritechNAPinChicago(Illinois),theSprintNAPinPennsauken(NewJersey)andthealreadyexistingMAEEast,operatedbyMCI

WorldCominVienna(Virginia).Inthecomingyearstheamountoftrafficstarted

toreduce,becauselargernetworksdecidedtosetupaprivateinterconnectionandthesmallernetworkswerelimitedtoregionalpublicinterconnections

Today,approximately90InternetExchangesarelocatedinNorthAmericaand

mostofthemareoperatedasacommercialservice.ThesecommercialInternet

Exchangeshaven’tbeenconfinedtowithintheAmericanborders,buthavereachedouttoEuropeandAsiaaswell.Non-commercialInternetExchangescan

alsobefoundinNorthAmerica,howevernumbersarefarless.HowmuchIXP

trafficisbeingexchangedwithinthisregionisalittleunclear,asmanyoftheIXPsdonotpubliclydisplaythesestatistics.About5%oftheworld’spopulation

islocatedinNorthAmericawithanInternetpenetration[5]ofover75%.TheUSmakesupabout13%ofthetotalnumberofInternetusersintheworld.

4.3 Asia

ThereisalargedifferencebetweenthecountrieswithinAsiawiththeirlevelofInternetaccessandconnectivity.ForinstancecountrieslikeChinaandJapanare

moredevelopedthancountrieslikeVietnam,Indonesia,NepalandIndia,where

theInternetpenetrationissignificantlylower.Whenyoulookatthepopulationstatistics[5]fromAsia,youcanseethat56.0%oftheworldpopulationislocated

inAsia.OfthetotalAsianpopulation,only21.5%hasaccesstoInternet.Inthelast10yearstheInternetgrowthhasincreasedtremendously,andthisseemsto

beacontinuingtrend.About42.0%ofthetotalInternetUsersintheworldare

locatedinAsia.ThiswouldsuggestthatalotoftrafficisgoingthroughtheAsianregions.

4.3.1 China

OneoftheveryfirstInternetlinks(64kbps)inHongKongwasin1991bytheChineseUniversityofHongKong(CUHK)[10].However,backinthosedaysit

wasstillunderBritishrule.AyearlaterallhighereducationalinstitutionsinHongKongwerelinkedtoeachotheroverthislink.TheycalledthisnetworkHARNET(HongKongAcademicandResearchNETwork).

In1994whenafewsmallISPswereestablishedtheInternetonthenon-academicsidestartedtogrowwhileservingmostlyindividualsubscribers.

TherealgrowthofInternetstartedin1995asmoreandmorecompaniesstarted

toseetheimportanceoftheInternet.Thiscausedamajorimpactontheavailabilityoflocalcontentandintra-HongKongcommunicationbecamemoreof

anecessity.HoweveralargenumberofISPsstillhadtorelyonotherISPsthathadalocalpresencetocompletetheredotransitforthem.

ThismassivegrowthofInternetcausedaneedforabetternetwork

infrastructuresoastolowerthecostsofoperations.ItwasnotveryeffectiveforISPstoroutetheirintra-HongKongtrafficoverseas,becausethelinkswere







20

4.3.4 Nepal

TheInternetmarketinNepal[13]isgrowingrapidly.Withlessthan1percentof

Nepal’spopulationusingtheInternetin2005,thisnumberisgrowingfast.AcompetitivemarketbetweenISPsandlowpricestoInternetaccessarethecause

ofthis.Rightnow31privateISPsofferInternettobusinessesandhome-users,

butonlytwoofthemdominatethemarketwithatotalshareofover70%.TheseISPsareWorldLinkandNepalTelecom.AnimportantsourcetoInternetaccess

fortheNepalesearecybercafés,itisbelievedtohaveoneofthehighestconcentrationofcybercafésintheworld.TheKathmanduValleyregionhasa

higherpenetrationofInternetaccess,becauseit’smoredevelopedwhilebeing

hardtogainaccesstothemountainregionswherelowincomespopulationscanbefound.TheNepaliWirelessNetworkingProjectisrunninganinitiativeto

makeInternetaccesspossibleintheruralpopulatedmountainregions.In2002thefirstInternetExchangewasestablishedinNepal,callednpIX.

Currentlytheyhavepeaktrafficofover100Mbpswith17connectedISPs,

includingthetwomostdominateISPs.NepalhasanInternetpenetration[5]of2.2%,andmakeupabout0.1%ofthetotalusersinAsia.

4.4 LatinAmerica

TheInternetpenetrationinLatinAmericaamongstitspopulationis39.5%.They

hold8.0%ofthetotalInternetusersintheworld.Howeverinthepast10years,

therehasbeenagiganticgrowth,inusers,soInternetisbecomingmoreandmorecommoninthisregion.OfalltheLatinAmericancountries,Brazilisbyfar

thelargestoftheInternetuserswithsome48.5%ofthecontinent’stotalandtheInternetuserpenetrationisataround37.8%.Thesecondbiggestcontributorto

InternetusageinLatinAmericaisArgentina,with17%followedbyColombiawith13.7%.However,theInternetpenetrationamongstthepopulationissignificantlyhigherinArgentina,comparedtoBrazil,with64.4%andColumbia

with48.7%respectively.

4.4.1 Brazil

In1962aresearchfoundationwasestablishedintheSaoPauloregioninBrazil,whichwasgiventhenameFAPESP(ResearchFoundationSaoPaulo)[14].In

1988theANSP(AcademicNetworkSaoPaulo)wascreatedasanacademic

networkbyFAPESP,whichwouldbethefirstBraziliannetworktobeintegrated

withtheInternetin1991andconnecteduniversitiesandresearchinstituteswithoneanother.TheANSPoperated‘NAPdoBrasil’,in2004Terremarkand

FAPESPdecidedtoworktogetherbylettingTerremarkoperatetheNAP,thiswouldbenefitbothpartiesconcernedandthegeneraldevelopmentofInternetin

Brazil.BesidesFAPESPanInternetSteeringCommittee(CGI.br)wascreatedinMay

1995,withtheaimofpromotingandimprovingtheparticipationofthesociety

towardstheimplementation,managementandusageoftheInternet.Oneoftheexecutivearms(CenterofStudyandResearchinNetworkTechnologyand

Operations,CEPTRO)oftheInternetSteeringCommitteeisresponsibleforprojectsandservicesthatareaimedtoimprovethequalityofInternetinBrazil.

OneoftheirprojectswasthecreationofPTTMetro,whicharetheBrazilian



21

InternetExchangePoints,

locatedinallofthemajorBraziliancities.Thisprojectwas

startedin2004.ThemostimportantfeaturesofthePTT

Metroaretobeneutralandindependentfromcommercialsuppliersandofferhighquality

service.Todaytheyhavesome

14IXPswithanaggregatedpeaktrafficofmorethan30Gbpsand

morethan200participants.TheInternetSteeringCommitteeis

responsibleforoperatingand

managingtheBrazilianInternetbackbone.

4.4.2 Argentina

In1989anassociationwas

formedby:ISPs,Datacenters,Solutions&Contentproviders

andmanyotherentities.This

associationwasnamedCABASEandcreatedNAPdeCABASEin

1997.ThepurposeofNAPde

CABASEwastoprovideaplacewhereallInternetplayers Figure5.4showstheLatinAmerican

couldpeerefficiently.Therewerea countriesthathaveoneormoreIXPs.fewthingstobeconsideredbeforethis Thewhitecountrieshavenoknown

couldbeaccomplished.Thelocation IXPsoperatingwithinthem.

ofthesitewasveryimportant.ItwaseventuallydecidedtolocateitinDowntownBuenosAires,thelargestandmost

importantcityinArgentinaandhasmanyInternetplayers.ThedecisionofthelocationwasalsobasedonthenetworkavailabilityforaccessingNAPde

CABASE.ThenextconsiderationwasthemodeloftheNAP.Itwasdecidedtouse

acooperativeandself-governedmodel.Anycostorinvestmentwouldbe

distributedamongstthemembersofNAPdeCABASE.Operatingagreementsaswellasproceduresaredevelopedandvotedonbythemembers.Eachmember

hastherightofonevote,regardlessofitssize.

NAPdeCABASEisalsocarrierneutral,soeventodaythesiteandinfrastructurearenotownedbyanyoftheirmember,whichminimizestheriskofbeing

affectedbyproblemsofamember.TheNAPhasspecialagreementsforeither

non-carrierornon-ISPmembers.Theycanonlyconnectiftheirbusinessesdon’taffectanyofthecurrentISPsorcarriersthatareconnected.Theaveragenumber

ofmembersconnectedtoNAPdeCABASEis35.SincethebeginningoftheNAP

multilateralpeeringwastheonlypeeringagreementspossibilitymadeavailable,butsincetwoyearsago,bilateralagreementshavebeenpermitted,butnot



22

activelypromoted.InNovember2010NAPdeCABASEoperatedthefirst

regionalNAPintheNeuquénProvince,startingwith7members.IthasaconnectiontotheNAPinDowntownBuenosAireswithaccesstoallthe

membersandtrafficthere,withthesameconditionsastheirmembers.ThiscreatesahugecostimprovementfortheArgentinaregion.Plansforother

regionalNAPshavealsobeenmade,howevernofurtherdetailshavecometohand.AfterthedeploymentofthefirstNAPdeCABASEagrowthinlocalhosting,contentande-commerceserviceswasseen,thisresultedinmorelocalcontent

becomingavailable.

4.5 Africa

InthelastcoupleofyearstheInternetusageinAfricahasgrowntremendously,

tosuchanextentthatthecurrentnetworkcapacitycan’tkeepupwiththedemand,whichactuallyledtofallingbehindontheirgrowth.Rightnowthey

couldbeconsideredasbeingalmost10yearsbehindonEuropeandNorthAmerica,howeverthisdifferenceonlylookssettogrowbiggerandbiggerover

thenextcoupleofyears.Inthelastcoupleofyearstheexistingconnectivitywas

mainlyoversatellites.Thisisveryexpensive,usuallyslowandsensitivetotrafficcongestion.

DependingonthepartoftheAfrican[15]regionthatyouarelookingat,the

growth,stateofInternetconnectivityandaccessisverydiverse.Somecountriesarewelldeveloped,whileothersareverypoorlydeveloped.Thismakesitvery

hardtoconnectallthecountriestoeachotherandprovidebetterInternetconnectivityandaccesswithintheAfricanregions.Oneofthemainreasonsfor

thepoorgrowthispoorinfrastructureandveryfragileend-to-endconnectivity

(forexample,connectivityfromthelocalISPtothehome-user),thisisincreasedbyalackofadequatenetworkbackuproutes,incasethepowergoesdownora

cableisbroken.OtherdisadvantagesforAfricaarelackofpower,highimporttaxes(togetroutersandswitchesintothecountry),lackofproperskilled

technicalstaff(theytendtoleavetomoredevelopedregionslikeNorthAmerica

orEurope,oncetheyareskilled),corruptionandmonopolies.InternationalcompaniesandorganizationshavestartedprojectstodeployproperInternet

connectivitytoalargernumberofpeoplethroughouttheregion.

Today,mostoftheAfricantrafficisroutedthroughEuropeorNorthAmericaevenifthedestinationoftherequestiswithinAfricanborders.Someofthe

trafficevenneedstogothroughbothEuropeandNorthAmericatothengoback

totheAfricanregion.HighinternationalbandwidthpricesarethebiggestcontributortoAfricanISPs,alotofthetrafficgoesoversatellitelinks,because

fibrelinksarelimitedandexpensivetodeploy.HightrafficcostsandthepoorconnectivitycanbeimprovedbybuildingregionalandnationalIXPs,thisalso

reducestheusageofinternationaltrafficlines.

ThefewcountriesinAfricathatdohaveaccesstointernationalfibredon’tfeelthebenefitsofusingthis,becausetheyarestillchargedatmonopolylevelprices

bystateownedoperators,whichstilldominatemostofAfricaandinmanyotherdevelopingregions,sosatellitelinksareusedinstead.Thisisbeginningto

changeasmanycountriesrevisetheircompetitionframeworksasexclusivities,

grantedtoincumbents,cometoanend.In2003therewere10countriesofthe53intheregionthathadanIXP,in2006thisnumberhadrisento16countries.



23

TodaythetotalnumberofIXPsinAfricais18andthisnumberlookstogrow

further.Asmoreandmoreroutesstabilise,morecustomerswillsubscribetoISPs

services.IntimethiswillresultinbetterQoSwithawiderimpactonthedeploymentofinfrastructureintheregions.Localpeoplearedemandingbetter

connectivityandtheISPshavenochoicebuttoinvest,soastomakemoremoney.Thisisgoodforthelocaleconomy.TodayfibrecablesarebeingdeployedrightacrossAfrica,howeverthisisalong

processwherealotcangowrong,suchassabotageonlandlinesthatare,either

brokenorstolen,whichkeepstheincumbentsfromdeployingfaster.Sometrytosolvethisbyproviding

wirelessaccesstotheirusers,butthishasother

implications,whichareout

ofthescopeofthisresearch.Inthenextcoupleofyears

theInternetconnectivity

shouldhaveimprovedfromwhereitistodayandit

shouldbemoreaccessibletoeverybody.

Atthismomentsome14%ofthetotalworldpopulation

[5]islocatedinAfrica,butonly10%ofthetotalAfrican

populationhasaccesstoInternet.Thismeansthereisalotofpotentialtogrowand

createlocalInternet

connectivitytohomeuserscomparedtothatofany

otherregionintheworld.

Figure5.5showstheAfricancountriesthathaveanIXP.Thewhitecountriesdonot

haveanIXP.SomeoftheAfricancountrieshavemorethan1IXP,howeverthisislimitedtojusttwocountries,EgyptenSouthAfrica.



24

5 IXPmodelsThereareapproximately300IXPs[16]aroundtheworld,howeverthereare

quiteanumberofdifferencesbetweentheseIXPs.NoIXPisthesame,theymay

lookquiteabitalike,butthingswillprobablybejustabitdifferentineachcase.ThissectionwilltakealookatthedifferencesbetweentheIXPmodels[16].

5.1 Commercialvs.non-commercial

AkeyfactortodifferentiatinganIXPiswhetheritiscommercialornot-for-profit.WhenanIXPiscommercialitsmaingoalistomakeaprofitandmore

oftenthannot,operatedbythesamecolocationcompanythathousesthem.OneoftheadvantagesofacommercialIXPisthattheyhavetheabilitytocharge

strategicdifferentialpricing,toattractmorecustomerstoconnecttothem.They

alsotendtoapproachtheirpricingbasedonwhatthemarketcanwithstand.DecisionsaboutthecommercialIXParemadebythecommercialIXoperatorand

notbythepartiesconnectedtothem.CommercialIXPsalsotendtocompetewith

eachotherinsteadofcooperating.ISPs,contentprovidersandotherpartiesconnectedtoacommercialIXarecalledcustomers,whileforanon-commercial

IXPthesesameparticipantsareusuallyknownasmembers.WhenanIXPisnon-commercial,alsoknownas“non-profit”(or“not-for-profit”)

themembersconnectedtotheIXPgenerallyhaveaninfluenceonthedecisions

thatmayaffectthefutureoftheIXP.Non-commercialIXPswereestablishedwiththeaimofassistingISPstosavecostsandtoimprovetheend-user’sInternet

connectivityspeed,insteadofbeingpurelyset-uptomakeaprofitandthisis

possiblywhathashelpedtheseIXPsbecomesosuccessful.

5.1.1 NeutralityTheneutralityofanIXPcanbeseenasanimportantfactorofitssuccessfulness.

Somethinkit’sthekeyfactor,whileothersseeitasimportant,butnot

necessarilythemainfactor.ThereisnoofficialdefinitionoftheneutralityofanIXP,thishastobedecidedbytheIXPitself.Forinstance,itcandecidetobe

carrierneutral,butnotcolocationneutralorpossiblybothcarrierandcolocation

neutral.Oristheneutralityseenasonlyorganizational?Thereisnorightorwrongwhenyouaretalkingabouttheneutrality,becauseitreallydependson

whatyouseeasbeingneutralforanIXP.Otherwaystostay‘neutral’arenotinterferingwithyourcustomersormembers,

howeveritmaybenecessarytoestablishrulesforyourcustomersor

membershipbaseandinformthemofwhattheycanandcannotdo.WhentheIXPdecidestoimplementorchangeitservices,doyoulistentoyourmembersor

doyoustay‘neutral’.Allofthismakesthedefinitionoftheneutralityhardto

define.

5.1.2 Ownershipandmanagement

WhenyoulookatthedifferencesinownershipofIXPsaroundtheworldyoucan

findquitesomedifferencesinthebodiesrunningthem.ThebodyoftheIXPisa

veryimportantfactor.ItdecideswhomakesthedecisionsabouttheIXPandhow

thesedecisionsaremade.FromanhistoricpointofviewthemostcommonbodyiswhenanacademicinstitutionownstheIXP.Thisiswheretheideaoriginally



25

camefrom,howeversometimestheownershipdevelopsintoamoreformalbody

overtheyears.WhenthisdevelopmentoccursitusuallyleadstoanIXPbecomingAssociationbased.WhereaBoardhasbeencreatedwhomakemajor

decisionsaboutthefutureoftheIXP.ItisverycommonthatthemembersofanassociationbasedIXPhavevotingpowersandhassomekindofinfluenceonthe

futureoftheirIXP.Anothercommonbodyisthebusinessunit.ThistypeofIXPsaimstomakeaprofitandusuallyprovidethecollocationspaceifthedecisionsmadebytheIXP

managementareaimedatmakingprofitwhichtheygenerallydobysellingas

muchrackspaceaspossibleandtendtosettheirpricesbasedonthemarket.

5.1.3 Fees

TherearealotofdifferencesbetweenfeesatIXPs.SomeIXPsevenclaimtobecompletelyfree,meaningthattheydonotchargeaportconnectionfeehowever

theystillmayrequiretheparticipantstopayforrackspaceandlocaltransportto

theIXP.Ingeneral,thisisdonebyveryfewIXPstoday.DependingonwhetheranIXPiscommercialornon-commercialmayalsoinfluencethelevelofthefees.

UsuallywhenanIXPisnon-commercialyouseeamonthlyfeeyouhavetopayforaccesstoyourport,sometimesyoualsohavetopayaconnectionor

establishmentfee.ThisisallreallydifferentperIXP.WhenlookingatcommercialIXPs,youcanseealotofdifferencesinpricingaswell.Oneofthereasonsforthis

isthatmostofthecommercialIXPscandotheirpricingbasedonthemarket.

Thismakesthemmoreflexibleinaway.SomeIXPswantyoutopayaone-timefee,andamonthlyfeeforyourport.Somegivetheportawayforfree,because

youarerequiredtorentrackspaceinoneoftheircolocationfacilities,or

dependentoncertainservicestheyoffer.AveryimportantfactorwhenlookingatthepricingbetweenIXPsiswhether

theyhavevolunteersworkingforthem,ortheyareproperlystaffed.Whenyouwanttobesureofsupport,thedecisionisusuallymadetochoosefordedicated

personnel.Itisalsoimportanttolookattheequipmentinthiscase,usuallywhen

anIXPisranbyvolunteers,theequipmentisusuallyfundedbydonationsfromcompanies,organizationsandindividuals,whileastaffedIXPprobablyhasmore

budgetforequipment.Whichmeanstheyarelessdependantonthirdparties.

5.1.4 Scopeofactivities

WhichservicesanIXPofferstoitsparticipantsisreallydependentonthescopeoftheiractivities.TheseactivitiesweredecideduponwhentheIXPstarted,or

whena‘Wewillseewhereitgoesfromthere’mentalitydidn’tworkanymore.

WhatexactlyisofferedbyanIXPisyetagaindependantonwhetheritiscommercialornon-commercial.CommercialIXPsusuallyprovidetheir

customerswithcolocation,hosting,security,customersupport,SLA’setc.All

theseservicesarealsousedtomakemoreprofit.Evenwhenlookingatnon-commercialIXPsthescopeofactivitiesalsotendsto

differsomewhat,fromjustofferingportstofullcustomersupport,SLA’sandrackspace.



26

5.1.5 Peeringagreementsandpolicies

Peeringcanbeachievedondifferent‘levels’,itcanbeimplementedinatechnical

way,oritcanalsobeimplementedinanon-technicalway,basedonthepeeringpoliciesfromconnectedIXPparticipants.Whenit’sbasedonthepeeringpolicies

oftheIXPparticipant,ithasnothingtodowiththeIXPitself,becauseit’spurely

anagreementbetweentwoormorepartiesthatconnecttotheIXPanddescribeshowtheywanttoexchangethetrafficandunderwhichconditionsthishastobe

done.

Therearetwowaystoimplementpeeringonatechnicalbasis.Thiscaneitherbe

doneviaprivatepeeringorthroughpublicpeeringwherethetrafficgoesoveranIXP.Adefinitionofbothcanbefoundinthetablebelow.

Privatepeering Physicaldirectconnectionbetweenonlytwoparties

Publicpeering(overIXP) Physicalconnectiontoamedium,managedbyathirdparty,through

whichmultiplepartiesareconnected

andapproachable

Table6.1:Differentwaystoimplementtechnicalpeering.

Theotherpossibilityofimplementingpeeringisinanon-technicalway,whereagreementsaremadebetweentwoormoreparties.Everyconnectedpartytoan

IXPhasaspecificpeeringpolicy,inwhichitdefineshowtheywanttoexchange

trafficandunderwhichconditions.Thedifferentpeeringpoliciesaredescribedinthetablebelow.Everyconnectedpartycanhaveoneofthesepolicies,based

ontheirbusinessmodelandpeeringconditions.Thispeeringpolicymaydifferperregion.ForexamplethelocalincumbentinacountryinEuropecanchooseto

havearestrictivepeeringpolicyinEurope,howeverwhenthisincumbent

expandsitbusinesstoadifferentregion,forexampleNorthAmerica,itmaydecidetohaveanopenpeeringpolicyasthismaymakemorefinancialsense.

TheconnectedpartiesatIXPsinNorthAmericamightbemoreinterestingtopeerwiththantheconnectedpartiesattheIXPinitshomecountry,wherethey

feelitwilldecreasetheirbusiness.

Policy Definition

OpenPeering Generallywillingtopeerwithanyone,withnoorfewprerequisites.

SelectivePeering Generallywillingtopeer,howevertherearesome

prerequisites,oncemetthisgenerallyleadstopeering.

RestrictivePeering Generallynotwillingtopeerextremelydifficulttomeetprerequisiteswithhighpossibilityofdenialoncetheyare

met.

NoPeering Nointentiontopeer.Trafficisexchangedviatransitagreements

Table6.2:PeeringpoliciesforIXPparticipants.

Theimplementationofthepeeringpoliciescanbedoneinanumberofdifferentways.Thishappensonanon-technicalwayandismoreagreementbasedonthe









30

5.4 Asia

TheAsianPacificregionisusuallythoughtofasaclosedcommunitywhereinterconnectingishardtomaintainandsetup.Thismaybetrue,howeverit’s

slowlybecomingeasiertointerconnectinAsia.Oneofthebiggestproblemsthe

Asianmarketisstrugglingwithistheartificiallyhighpriceoftransit,transatlanticcostsandinterconnectingcostsatIXPs.MostofthedominantISPs

stillthink:‘foryoutogetmoreroutestotheInternetyouhavetobuytransitfromme’.Thisisonlypartlytrue.ItworksfortheISPs,howeveritdoesn’thelp

achievegettingbetterconnectivityinthecountrynortheregion.Theideaof

peeringdoesnotinterestthem,eitherduetothefactthattheydon’tknowthebenefitsofconnectingatanIXPorareafraidoflosingbusiness.Thiskeepsthe

pricesofInternettrafficveryhighforsmallerISPsandoverseascompanies.

CreatinganIXPwouldbeeasy,howeveralotofpartiesdon’tseethebenefitfromdoingthis,becauseofthetremendoushighpricesofconnectingtoanIXP.

Asiaisalsotroubledwithalotofregulatedtelecoms.Pricesinde-regulatedregionsaregenerallylowerthanregulatedregions.ManycountriesinAsiaare

dependantonneighbouringcountriesfortransitconnectivity.Anotherproblem

inAsiaareculturedifferences,Tohelpsolvealloftheabovementionedproblemsitsimportantforallinvolvedpartiestoparticipateinsuchforumsas

APNIC,SANOGandApricot.Thishelpstocreateabetterunderstandingofthe

conceptofIXPsandpeering,butalsogivespeopletheopportunitytoexchangeideasandtalkaboutproblemswithmoreexperiencedpeoplefromalloverthe

world.It’salsoagoodsourceforfuturecustomersandpeeringrelations.Bothcommercialasnon-profitIXPscanbefoundthroughoutAsia.However,

mostofthemorecommercialIXPscanbefoundinwealthierregions,likeChina,

JapanandAustralia,whilelessdevelopedregionslikeIndia,NepalandthePhilippinestendtohavemorenon-profitIXPs.

5.5 LatinAmerica

MostoftheIXPsinLatinAmericaareofanon-profitnature.TheownershipoftheIXPvariesfromgovernmenttoassociationbasedtoanorganizationof

telecomoperators.Mostcountriesareheavilyregulatedbytheirgovernments.

Theseregulationsmakeitharderforcompanies(outsideofLatinAmerica)toconnecttoanIXP,orforlocalbusinesstoconnecttoanIXPinadifferentcountry.

Thisholdsbacktheregionfrommakingfasterprogress.AlmosteverycountryinLatinAmericahasanIXP,BrazilhasalargenumberofIXPsandarealloperated

byCGI.br.InArgentinaNAPdeCABASEoperatestheIXP.SomeIXPsinLatinAmericaforcealltheirparticipantstopeerwitheachotherto

encouragetrafficontheirplatform,whileotherleavethisoptiontotheirparticipants.Whenpeeringisforced,(international)companiestendtobeless

willingtopeerattheseIXPs.ThegrowthoftheLatinAmericanInternet

ecosystemwillmostprobablygrowfaster,onceoneormoreinternationalcompanieshaveestablishedthemselvesinthisregionandareinterestingtopeer

with,thiswillattractevenmoreinternationalbusinesses.Thiswillalsomakeit

easiertoexpandtootherneighboringcountries;whichhelpstheirInternetecosystemsgrowevenfurther.CurrentlylargeproportionsoftheLatinAmerican

















38

culturalneedsandmore.Allofthismakesitalotharderforacompanytoexpand

toanewcountry.AscanbeseeninFigure7.2(MapoftheU.S.stateswithanIXP),therearealarge

numberofstatesthathavenoIXPsatall.Thismeansthatthephysicallocationbetweencities(inthestates)thatalreadyhaveanIXP(theoneswithoutare

probablylessinterestingforacompanytoestablishthemselvesanddobusiness)canbequitelarge,especiallycomparedtoEurope,wheremostofthelargercitieshaveatleastoneIXP.Thismakesitmoreattractiveforcompaniesor

organizationstoexpandtodifferentcountries.Thecostsoffibrewillprobablybe

lessexpensive,becausethedistancebetweentwocitiesisrelativelyclose,comparedwiththoseintheU.S.Thismakesitmoreexpensiveforcompaniesto

reachouttoanIXP,letalonemultipleIXPs.Especiallyforsmallercompaniesit’scheapertobuytransitinsteadofconnectingtoanIXP.TheaddedvalueoftheIXP

won’tbringenoughrewardtotheirbusiness.Whentwocompaniesdodecideto

exchangetrafficinthesamedatacenter,thisisusuallydoneviacrossconnects,becausetheportpricesaremoreexpensivethanthephysicalcablelaidbetween

theirtwonetworks(thiscableisusuallyjustwithinthedatacenteritself,sonot

soexpensivecomparedtorentingaportattheIXP).Largequantitiesoftrafficgoingbetweentwonetworksoftenmakesacrossconnectsmoreoftenandmore

financiallyinteresting,butofferingtheIXPservicecanactasagreatdrawcard.







41

Acknowledgements

Specialthankstoeverybodywhotooktimeintoansweringmy,oftenmany

(annoying)questions,requestsande-mailsingeneral.Youhaveallbeenveryuseful!Icouldwritedownalistofallpeoplewhohavehelpedmeouthere,with

theriskofforgettingnamesormakingpeopleupsetiftheydonotwishtobenamed.

Peoplefromthefollowingcompanies,organizationsandForumshavecontributedtothisdocument:

- 17thEuro-IXForumattendees- NANOG50attendeeswhowerewillingtosharetheirbusinesscards

- DiversenumberofpeopleparticipatingintheAsianIXPscene- Akamai

- Amazon- APNIC- Equinix

- HurricaneElectric

- IPLAN- JPIX

- JPNAP- LACNIC

- Limelight

- PhoenixNAP- RipeNCC

- Telx- Yahoo

OfcoursealsoallthepeoplethatIforgottomentionabove.Allyourhelphasbeenverymuchappreciated!





43

[15] W. Stucke, Challenges facing ISPs in Africa: a view from an African ISP.

2008, http://www.afrispa.org/dpages/Powerpoint/Challenges%20facings

%20ISPs%20in%20Africa.ppt

[16] EuropeanInternetExchangeAssociation,Euro-IX,http://www.euro-

ix.net/(seenotebelow)

[17] S.Radovcic,EuropeanInternetExchangeAssociation2010ReportonEuropeanIXPs,2010http://www.euro-ix.net/resources/reports/euro-

ix_report_2010.pdf

[18] B.Norton,JapanInternetPeeringEcosystem,2005http://peering.drpeeri

ng.net/white-papers/Ecosystems/Japan-Peering-Ecosystem.php

[19] MinistryofInternalAffairsandCommunications,Understandingofthe

totalInternettrafficinourcountry,2010http://www.soumu.go.jp/main_ content/000080807.pdf(Japanese)

[20] G.Huston,DistributionReports,ASResourceAllocations,2010

http://resources.potaroo.net/iso3166/archive/

Note: Some statistics from Euro-IX are only accessible with login credentials.







3

2009

Country

Unique

ASNs

Allocated

ASNs

Advertised

ASNs

% ASNs at

IXPs

Austria 111 335 274 40.51%

Belgium 62 146 122 50.82%Croatia 15 75 64 23.44%

Cyprus 6 54 39 15.38%

Czech Republic 93 396 275 33.82%

Denmark 50 187 134 37.31%

Estonia 18 28 23 78.26%

Finland 28 143 114 24.56%

France 234 602 442 52.94%

Germany 470 1259 963 48.81%

Greece 15 145 105 14.29%

Hungary 47 196 148 31.76%

Iceland 15 29 27 55.56%Ireland 47 105 83 56.63%

Italy 135 578 464 29.09%

Latvia 158 207 157 100.64%

Luxembourg 16 29 24 66.67%

Malta 0 23 17 0.00%

Netherlands 447 458 340 131.47%

Norway 72 151 101 71.29%

Poland 110 1035 884 12.44%

Portugal 20 64 54 37.04%

Romania 69 1192 883 7.81%

Russia 400 2918 2435 16.43%Slovakia 47 79 63 74.60%

Slovenia 16 176 150 10.67%

Spain 60 324 253 23.72%

Sweden 102 421 300 34.00%

Switzerland 159 450 345 46.09%

Ukraine 161 1551 1173 13.73%

United Kingdom 493 1760 1263 39.03%

Total 3676 15116 11719 41.25%





Growth2007-2010

Country

Unique

ASNs

Allocated

ASNs

Advertised

ASNs

% ASNs at

IXPs

Austria 21.11% 18.12% 18.78% 1.97%

Belgium -16.46% 35.34% 42.70% -41.45%Croatia 0.00% 50.00% 60.98% -37.88%

Cyprus 0.00% 29.79% 39.29% -20.51%

Czech Republic 38.89% 386.13% 304.35% -65.65%

Denmark 0.00% 48.59% 58.76% -12.99%

Estonia 5.88% 56.52% 55.00% -31.69%

Finland 15.38% 30.58% 39.33% -17.18%

France 19.09% 38.48% 47.72% -19.38%

Germany 61.64% 12.30% 16.30% 38.98%

Greece -20.00% 6.38% 2.15% -21.68%

Hungary 10.64% 13.14% 32.77% -16.67%

Iceland 7.14% 59.09% 82.35% -41.24%Ireland 72.73% 40.74% 50.91% 14.46%

Italy 85.39% 23.90% 28.61% 44.15%

Latvia 42.37% 44.00% 42.98% -0.43%

Luxembourg 84.62% 94.44% 100.00% -7.69%

Malta 0.00% 4.17% -25.00% 0.00%

Netherlands 52.97% 26.77% 31.72% 16.13%

Norway 17.46% 52.83% 58.57% -25.93%

Poland 142.19% 63.44% 55.78% 55.46%

Portugal 5.26% 8.47% 37.50% -23.44%

Romania 112.20% 1.54% -14.46% 148.06%

Russia 125.73% 76.38% 83.56% 22.97%Slovakia 20.45% 37.70% 50.00% -19.70%

Slovenia 6.67% 85.45% 82.42% -23.53%

Spain 26.92% 40.78% 45.26% -12.63%

Sweden 41.46% 29.91% 39.66% 1.29%

Switzerland 50.45% 49.59% 39.78% 7.63%

Ukraine 114.81% 64.44% 57.86% 36.08%

United Kingdom 24.82% 28.77% 33.64% -6.60%

Total 37.74% 50.25% 51.59% -1.91%

Note:Statisticsfromthecountriesindicatedingreenwerenotpresentfor

certainyears,thiscountsfor:Cyprus,Denmark,MaltaandSlovenia.Thecalculationshavebeenmadefromthefirstyeardatawasavailableforthat

specificcountry.

894 Ixp Research PDF

Documents