PATHVECTORROUTINGANDTHEBORDERGATEWAYPROTOCOL

READING:SECTIONS4.3.3PLUSOPTIONALREADING

COS461:ComputerNetworksSpring2010(MW3:00‐4:20inCOS105)

MikeFreedmanhDp://www.cs.princeton.edu/courses/archive/spring10/cos461/

1

GoalsofToday’sLecture

•  Path‐vectorrouQng– Fasterconvergencethandistancevector– MoreflexibilityinselecQngpaths

•  InterdomainrouQng– AutonomousSystems(AS)– BorderGatewayProtocol(BGP)

•  BGPconvergence– CausesofBGProuQngchanges– PathexploraQonduringconvergence

2

InterdomainRouQngandAutonomousSystems(ASes)

3

InterdomainRouQng•  InternetisdividedintoAutonomousSystems– DisQnctregionsofadministraQvecontrol–  Routers/linksmanagedbyasingle“insQtuQon”–  Serviceprovider,company,university,…

•  HierarchyofAutonomousSystems–  Large,Qer‐1providerwithanaQonwidebackbone– Medium‐sizedregionalproviderwithsmallerbackbone–  Smallnetworkrunbyasinglecompanyoruniversity

•  InteracQonbetweenAutonomousSystems–  InternaltopologyisnotsharedbetweenASes– …but,neighboringASesinteracttocoordinaterouQng

4

AutonomousSystemNumbers

5

AS Numbers are 16 bit values.

•  Level 3: 1 •  MIT: 3 •  Harvard: 11 •  Yale: 29 •  Princeton: 88 •  AT&T: 7018, 6341, 5074, … •  UUNET: 701, 702, 284, 12199, … •  Sprint: 1239, 1240, 6211, 6242, … •  …

Currently over 50,000 in use.

whois–hwhois.arin.netas88

6

OrgName: Princeton University OrgID: PRNU Address: Office of Information Technology Address: 87 Prospect Avenue City: Princeton StateProv: NJ PostalCode: 08540 Country: US

ASNumber: 88 ASName: PRINCETON-AS ASHandle: AS88 Comment: RegDate: Updated: 2008-03-07

RTechHandle: PAO3-ARIN RTechName: Olenick, Peter RTechPhone: +1-609-258-6024 RTechEmail: polenick@princeton.edu …

ASNumberTrivia

•  ASnumberisa16‐bitquanQty–  So,65,536uniqueASnumbers

•  Somearereserved(e.g.,forprivateASnumbers)–  So,only64,510areavailableforpublicuse

•  ManagedbyInternetAssignedNumbersAuthority– Givesblocksof1024toRegionalInternetRegistries–  IANAhasallocated39,934ASnumberstoRIRs(Jan’06)

•  RIRsassignASnumberstoinsQtuQons–  RIRshaveassigned34,827(Jan’06)– Only21,191arevisibleininterdomainrouQng(Jan’06)

•  Recentlystartedassigning32‐bitAS#s(2007)7

GrowthofASnumbers

8

InterdomainRouQng

•  AS‐leveltopology– DesQnaQonsareIPprefixes(e.g.,12.0.0.0/8)– NodesareAutonomousSystems(ASes)– EdgesarelinksandbusinessrelaQonships

9

1

2

3 4

5

6 7

Client Web server

ChallengesforInterdomainRouQng

•  Scale– Prefixes:200,000,andgrowing– ASes:20,000+visibleones,and60Kallocated– Routers:atleastinthemillions…

•  Privacy– ASesdon’twanttodivulgeinternaltopologies– …ortheirbusinessrelaQonshipswithneighbors

•  Policy– NoInternet‐widenoQonofalinkcostmetric– Needcontroloverwhereyousendtraffic– …andwhocansendtrafficthroughyou

10

Path‐VectorRouQng

11

Shortest‐PathRouQngisRestricQve•  Alltrafficmusttravelonshortestpaths

•  AllnodesneedcommonnoQonoflinkcosts

•  IncompaQblewithcommercialrelaQonships

12

Regional ISP1

Regional ISP2

Regional ISP3

Cust1 Cust3 Cust2

National ISP1

National ISP2

YES

NO

Link‐StateRouQngisProblemaQc

•  TopologyinformaQonisflooded– Highbandwidthandstorageoverhead– ForcesnodestodivulgesensiQveinformaQon

•  EnQrepathcomputedlocallypernode– Highprocessingoverheadinalargenetwork

•  MinimizessomenoQonoftotaldistance– Worksonlyifpolicyissharedanduniform

•  TypicallyusedonlyinsideanAS– E.g.,OSPFandIS‐IS

13

DistanceVectorisontheRightTrack

•  Advantages– Hidesdetailsofthenetworktopology– Nodesdetermineonly“nexthop”towardthedest

•  Disadvantages– MinimizessomenoQonoftotaldistance,whichisdifficultinaninterdomainsemng

– SlowconvergenceduetothecounQng‐to‐infinityproblem(“badnewstravelsslowly”)

•  Idea:extendthenoQonofadistancevector– Tomakeiteasiertodetectloops

14

Path‐VectorRouQng

•  Extensionofdistance‐vectorrouQng–  SupportflexiblerouQngpolicies– Avoidcount‐to‐infinityproblem

•  Keyidea:adverQsetheenQrepath– Distancevector:senddistancemetricperdestd–  Pathvector:sendtheen,repathforeachdestd

15

3 2 1

d

“d: path (2,1)” “d: path (1)”

data traffic data traffic

FasterLoopDetecQon

•  Nodecaneasilydetectaloop– LookforitsownnodeidenQfierinthepath– E.g.,node1seesitselfinthepath“3,2,1”

•  Nodecansimplydiscardpathswithloops– E.g.,node1simplydiscardstheadverQsement

16

3 2 1

“d: path (2,1)” “d: path (1)”

“d: path (3,2,1)”

FlexiblePolicies

•  Eachnodecanapplylocalpolicies–  PathselecQon:Whichpathtouse?

–  Pathexport:WhichpathstoadverQse?

•  Examples– Node2maypreferthepath“2,3,1”over“2,1”

– Node1maynotletnode3hearthepath“1,2”

17

2 3

1

2 3

1

BorderGatewayProtocol(BGP)

18

BorderGatewayProtocol

•  InterdomainrouQngprotocolfortheInternet

– Prefix‐basedpath‐vectorprotocol– Policy‐basedrouQngbasedonASPaths– Evolvedduringthepast18years

19

•  1989 : BGP-1 [RFC 1105], replacement for EGP

•  1990 : BGP-2 [RFC 1163] •  1991 : BGP-3 [RFC 1267] •  1995 : BGP-4 [RFC 1771], support for CIDR •  2006 : BGP-4 [RFC 4271], update

BGPOperaQons

20

Establish session on TCP port 179

Exchange all active routes

Exchange incremental updates

AS1

AS2

While connection is ALIVE exchange route UPDATE messages

BGP session

IncrementalProtocol

•  AnodelearnsmulQplepathstodesQnaQon– StoresalloftheroutesinarouQngtable– AppliespolicytoselectasingleacQveroute– …andmayadverQsetheroutetoitsneighbors

•  Incrementalupdates– Announcement

•  UponselecQnganewacQveroute,addnodeidtopath•  …and(opQonally)adverQsetoeachneighbor

– Withdrawal•  IftheacQverouteisnolongeravailable•  …sendawithdrawalmessagetotheneighbors

21

IncrementalProtocol

•  AnodelearnsmulQplepathstodesQnaQon– StoresalloftheroutesinarouQngtable– AppliespolicytoselectasingleacQveroute– …andmayadverQsetheroutetoitsneighbors

•  Incrementalupdates– Announcement

•  UponselecQnganewacQveroute,addnodeidtopath•  …and(opQonally)adverQsetoeachneighbor

– Withdrawal•  IftheacQverouteisnolongeravailable•  …sendawithdrawalmessagetotheneighbors

22

ExportacQveroutes

•  InconvenQonalpathvectorrouQng,anodehasonerankingfuncQon,whichreflectsitsrouQngpolicy

23

BGPRoute•  DesQnaQonprefix(e.g.,128.112.0.0/16)•  RouteaDributes,including

–  ASpath(e.g.,“701888”)–  Next‐hopIPaddress(e.g.,12.127.0.121)

24

AS 88 Princeton

128.112.0.0/16 AS path = 88 Next Hop = 192.0.2.1

AS 7018 AT&T

AS 11 Yale

192.0.2.1

128.112.0.0/16 AS path = 7018 88 Next Hop = 12.127.0.121

12.127.0.121

ASPATHADribute

25

AS7018 128.112.0.0/16 AS Path = 88

AS 1239 Sprint

AS 1755 Ebone

AT&T

AS 3549 Global Crossing

128.112.0.0/16 AS Path = 7018 88

128.112.0.0/16 AS Path = 3549 7018 88

AS 88

128.112.0.0/16 Princeton

Prefix Originated

AS 12654 RIPE NCC RIS project

AS 1129 Global Access

128.112.0.0/16 AS Path = 7018 88

128.112.0.0/16 AS Path = 1239 7018 88

128.112.0.0/16 AS Path = 1129 1755 1239 7018 88

128.112.0.0/16 AS Path = 1755 1239 7018 88

BGPPathSelecQon

•  Simplestcase– ShortestASpath– ArbitraryQebreak

•  Example– Three‐hopASpathpreferredoverafive‐hopASpath

– AS12654preferspaththroughGlobalCrossing

•  But,BGPisnotlimitedtoshortest‐pathrouQng– Policy‐basedrouQng

26

128.112.0.0/16 AS Path = 3549 7018 88

AS 1129

128.112.0.0/16 AS Path = 1129 1755 1239 7018 88

AS 3549 Global Crossing

AS 12654 RIPE NCC RIS project

Global Access

ASPathLength!=RouterHops

•  ASpathmaybelongerthanshortestASpath

•  Routerpathmaybelongerthanshortestpath

27

2 AS hops, 8 router hops

s d

3 AS hops, 7 router hops

BGPConvergence

28

CausesofBGPRouQngChanges

•  Topologychanges–  Equipmentgoingupordown– Deploymentofnewroutersorsessions

•  BGPsessionfailures– Duetoequipmentfailures,maintenance,etc.– Or,duetocongesQononthephysicalpath

•  ChangesinrouQngpolicy–  Changesinpreferencesintheroutes–  Changesinwhethertherouteisexported

•  PersistentprotocoloscillaQon–  ConflictsbetweenpoliciesindifferentASes

29

BGPSessionFailure•  BGPrunsoverTCP

–  BGPonlysendsupdateswhenchangesoccur

–  TCPdoesn’tdetectlostconnecQvityonitsown

•  DetecQngafailure–  Keep‐alive:60seconds–  HoldQmer:180seconds

30

AS1

AS2

•  ReacQngtoafailure–  Discardallrouteslearnedfromtheneighbor–  Sendnewupdatesforanyroutesthatchange–  Overheadincreaseswith#ofroutes

• WhymanyQer‐1ASesfilterprefixes</24

RouQngChange:BeforeandAter

31

0

1 2

3

0

1 2

3

(1,0) (2,0)

(3,1,0)

(2,0)

(1,2,0)

(3,2,0)

RouQngChange:PathExploraQon

•  AS1– Deletetheroute(1,0)–  Switchtonextroute(1,2,0)–  Sendroute(1,2,0)toAS3

•  AS3–  Sees(1,2,0)replace(1,0)–  Comparestoroute(2,0)–  SwitchestousingAS2

32

0

1 2

3

(2,0)

(1,2,0)

(3,2,0)

RouQngChange:PathExploraQon

•  IniQalsituaQon– DesQnaQon0isalive– AllASesusedirectpath

•  WhendesQnaQondies– AllASeslosedirectpath– Allswitchtolongerpaths–  Eventuallywithdrawn

•  E.g.,AS2–  (2,0)(2,1,0)–  (2,1,0)(2,3,0)–  (2,3,0)(2,1,3,0)–  (2,1,3,0)null 33

1 2

3

0

(1,0) (1,2,0) (1,3,0)

(2,0) (2,1,0) (2,3,0)

(2,1,3,0)

(3,0) (3,1,0) (3,2,0)

BGPConvergesSlowly

•  Pathvectoravoidscount‐to‐infinity–  But,ASessQllmustexploremanyalternatepaths– …tofindthehighest‐rankedpaththatissQllavailable

•  Fortunately,inpracQce– MostpopulardesQnaQonshaveverystableBGProutes– AndmostinstabilityliesinafewunpopulardesQnaQons

•  SQll,lowerBGPconvergencedelayisagoal–  Canbetensofsecondstotensofminutes– HighforimportantinteracQveapplicaQons– …orevenconvenQonalapplicaQon,likeWebbrowsing

34

35

BGPNotGuaranteedtoConverge

Exampleknownasa“disputewheel”

(3d)isavailable

(2d)isavailable

(3d)isnotavailable

(1d)isavailable (2d)isnot

available

(1d)isnotavailable

Conclusions

•  BGPissolvingahardproblem–  RouQngprotocoloperaQngataglobalscale– Withtensofthousandsofindependentnetworks–  Thateachhavetheirownpolicygoals– Andallwantfastconvergence

•  KeyfeaturesofBGP–  Prefix‐basedpath‐vectorprotocol–  Incrementalupdates(announcementsandwithdrawals)

•  Nextlecture:Tricksforsemngpolicy!

36