Post on 31-Dec-2020
IPv6prefixassignmentforend-customers- persistentvs non-
persistent,andwhatsizetochoose.
(BestCurrentOperationalPracticeforoperators)JanŽorž,InternetSociety/Zavod Go6
Whatisthisdocumentallabout?• ThisdocumentdiscussesthemainissuesrelatedtotheoperationalpracticesfortheassignmentofIPv6prefixesforend-customers.
• MakingwrongchoiceswhendesigningyourIPv6networkwillsoonerorlaterhavenegativeimplicationsonyourdeploymentandrequirefurthereffortsuchasrenumberingwhenthenetworkisalreadyinoperation.Thetemptationtotake“easy”approachesforquickerdeploymentshouldthereforeberesisted.
Co-authors• JanŽorž <zorz@isoc.org>,• SanderSteffann <sander@steffann.nl>,• Primož Dražumerič <Primoz.Drazumeric@telekom.si>,• MarkTownsley <townsley@cisco.com>,• AndrewAlston<andrew.alston@liquidtelecom.com>,• Gert Doering <gert@space.net>,• Jordi Palet <jordi.palet@consulintel.es>,• JenLinkova <furry@google.com>,• LuisBalbinot <lbalbinot@brfibra.com.br>,• KevinMeynell <meynell@isoc.org>,• LeeHoward<lee.howard@retevia.net>
Draftv.1
• https://sinog.si/docs/draft-IPv6pd-BCOP-v1.pdf• [root@webserver]#grep draft-IPv6pd-BCOP-v1.pdfsinog-ssl_access_log*|wc -l
909
Draftv.2
• https://sinog.si/docs/draft-IPv6pd-BCOP-v2.pdf• Numberofcommentsandsuggestionson-listandoff-list…
• PresentedandgatheredsomecommentsalsoatRIPEBCOPTFmeetingonMonday
• Majorityofco-authors,presentatRIPE74meetinginBudapestgatheredinalobbybaronTuesdaytodotheeditorialcycle,followedbylanguagepass.
• SenttoRIPEIPv6mailingliston11th Mayat02:44am
Editingdraftv.2
Agenericsetofrecommendations:
a)IPv6isnotthesameasIPv4.InIPv6youassignanumberof“n”/64prefixestoeachend-customersite,sotheyareabletohaveasmanysubnetsastheywish.YoushouldnotbeconcernedwithexhaustingtheIPv6addressingspace,andyoushouldthinkbigwhenplanningfuturerequirements.Ifyouneedmorespace,youcangobacktoyourRegionalInternetRegistryandprovidingyouraddressingplanjustifiesit,youcanobtainmoreIPv6addresses.
Agenericsetofrecommendations:b)Ifyouwantasimpleaddressingplan,youshouldconsiderthesethreeoptions:1)/48foreachend-customer.ThiswillworkverywellforcustomerscomingfromotherISPs,thosethathavetheirownULA,orhavebeenusingtransitionmechanisms.Thiswillalsobeeasierwhenyouhaveamixofcustomersusingthesameinfrastructure,whethertheyareresidentialcustomers,SMEsorevenlargecorporates.2)Differentiateamongsttypesofcustomers,evenifthiswillincreasethecomplexityofyournetworkandthoseofyourcustomers,byoffering/48forbusinesscustomersand/56forresidentialcustomers.3)Atrade-offamongsttheprevioustwooptionsbyreservinga/48forresidentialcustomers,butactuallyjustassigningthemthefirst/56.Thereaspecificcaseforcellularphonestobeassignedasingle/64pereachPDPcontext,butthisisoutofscopeofthisdocument.
Agenericsetofrecommendations:
c)Inordertofacilitatetroubleshootingandhaveafutureproofnetwork,youshouldconsidernumberingtheWANlinksusingGUAs(GlobalUnicastAddresses),usinga/64prefixoutofadedicatedpoolofIPv6prefixes.Ifyoudecidetouse/127foreachpoint-to-pointlink,itisadvisabletoallocatea/64foreachlinkandjustuseone/127outofit.
Agenericsetofrecommendations:
d)Non-persistentprefixesareconsideredharmfulinIPv6asyoucan’tavoidissuesthatmaybecausedbysimpleend-customerpoweroutages,soassigningpersistentprefixesisasaferandsimplerapproach.Furthermore,thisavoidstheneedforexpensivelogging,increasesyourchancestooffernewbusinesstocustomers,anddecreasesyourcustomerchurn.
Acknowledgements
TheauthorswouldliketothankNathalieKunneke-Trenaman,MikaelAbrahamsson,JasonFesler,MartinLevy,IanDickinson,PhilipHomburg, IvanPepelnjak,MatthiasKluth,Ondřej Caletka,NickHilliard,PaulHoffmanandRomanNurul Islamforconstructivecritic,suggestionsandideashowtomakethisdocumentbetter.SpecialthanksgotoRIPEIPv6WorkingGroupcommunityandChairsforacceptingthisdocumentfortechnicalreview,andalsotheRIPEBCOPTaskForcecommunityandChairsforensuringitdoesconformwithactualbestoperational practice.
Q&A
Suggestions?Comments?
Ideas?
Wayforward?