BGP: Introduction and Issues

Advanced Networks 1

BGP: Introduction and Issues

Advanced Networks 2

What Is BGP?

Border Gateway Protocol BGP-4Border Gateway Protocol BGP-4

The de-facto interdomain routing protocolThe de-facto interdomain routing protocol

BGP includes specifications:BGP includes specifications:• Which information gets advertised and howWhich information gets advertised and how

BGP includes a routing protocol: BGP includes a routing protocol: • Establishes and uses a routing tableEstablishes and uses a routing table

Internal Gateway Protocol (I-BGP in the book)Internal Gateway Protocol (I-BGP in the book)

Advanced Networks 3

Why Is There Such Fuss about BGP?

BGP dictates routing at the AS levelBGP dictates routing at the AS level• Absence of understanding: poor performanceAbsence of understanding: poor performance

BGP is complicatedBGP is complicated• Designed to be flexibleDesigned to be flexible• Involves multiple fieldsInvolves multiple fields

Understanding BGP behavior is not intuitiveUnderstanding BGP behavior is not intuitive• Implementation and business policiesImplementation and business policies

The routing of the Internet relies on BGPThe routing of the Internet relies on BGP

Advanced Networks 4

Some Open Questions

How well does BGP work now?How well does BGP work now?

How can I manage a BGP network?How can I manage a BGP network?

How secure and robust is BGP?How secure and robust is BGP?• Cyber-terrorismCyber-terrorism

How would we re-design BGP now?How would we re-design BGP now?

How well will BGP scale for our future How well will BGP scale for our future needs?needs?

Advanced Networks 5

Roadmap

Introduction to BGPIntroduction to BGP

Highlights of BGP issuesHighlights of BGP issues

Goal: instigate interest in BGPGoal: instigate interest in BGP

Advanced Networks 6

Some Basic Numbers

• 17,000 Autonomous Systems approx.17,000 Autonomous Systems approx.• Corporate NetworksCorporate Networks• ISP Internal NetworksISP Internal Networks• National Service ProvidersNational Service Providers

• Identified by ASN a 16 bit valueIdentified by ASN a 16 bit value• Assigned by IANAAssigned by IANA

• Superlinear growth (Huston, Siganos et al.Superlinear growth (Huston, Siganos et al.))

Advanced Networks 7

How A BGP graph Looks Like

Each AS has Each AS has designated BGP designated BGP routersrouters

BGP routers of an BGP routers of an AS communicate AS communicate internally with internally with another protocol another protocol (IGP)(IGP)

AS 1

AS 2

AS 3

AS 4

AS 5

Advanced Networks 8

IP Addresses and Prefixes

IP addresses have 32 bits: 4 octets of bits IP addresses have 32 bits: 4 octets of bits (IPv4)(IPv4)

A prefix is a group of IP addressesA prefix is a group of IP addresses

128.32.101.5 is an IP address (32 bits)128.32.101.5 is an IP address (32 bits)

128.32.0.0/16 is a prefix of the 16 first bits:128.32.0.0/16 is a prefix of the 16 first bits:• 128.32.0.0 – 128.32.255.255 (2^16 addresses)128.32.0.0 – 128.32.255.255 (2^16 addresses)

128.32.4.0/24 is a prefix of the 24 first bits - 128.32.4.0/24 is a prefix of the 24 first bits - longerlonger

Advanced Networks 9

Routing is Based on Prefixes

A BGP Routing table has prefixes for entriesA BGP Routing table has prefixes for entries

For a IP address of a packet, find For a IP address of a packet, find longest matchlongest match

Example: packet IP 128.32.101.1Example: packet IP 128.32.101.1

128.1.1.4 matches the first 8 bits – no match!128.1.1.4 matches the first 8 bits – no match!

128.32.0.0/16 match for 16 bits128.32.0.0/16 match for 16 bits

128.32.101.0/24 is a longer match 128.32.101.0/24 is a longer match

Advanced Networks 10

Prefix Matching in More Detail

For a IP address of a packet, find For a IP address of a packet, find longest matchlongest matchExample: Compare Example: Compare • packet IP 128.32.101.1packet IP 128.32.101.1• With 128.32.0.0/16With 128.32.0.0/16• IP : 01000000. 001000000. 01100101 .00000001IP : 01000000. 001000000. 01100101 .00000001• Mask : 11111111. 111111111. 00000000 .00000000Mask : 11111111. 111111111. 00000000 .00000000• AND : 01000000. 001000000. 00000000 .00000000AND : 01000000. 001000000. 00000000 .00000000• Prefix : 01000000. 001000000. 00000000. 00000000Prefix : 01000000. 001000000. 00000000. 00000000• Equal? YesEqual? Yes


Advertising Routing Information

Each AS advertises what it can reach from Each AS advertises what it can reach from each BGP routereach BGP router

Policies I: filter what you advertisePolicies I: filter what you advertise

Policies II: filter from what you hear Policies II: filter from what you hear advertised advertised

Build up a BGP routing tableBuild up a BGP routing table• Remember which prefix you hear from which linkRemember which prefix you hear from which link


What Does a Routing Table Look Like?

Origin AS “owns” the addressOrigin AS “owns” the addressRouting tables can have peculiaritiesRouting tables can have peculiarities

Prefix Origin AS Path

128.32.0.0/16 123 14 56 123

123 34 101 203 123

128.32.101.0/24 15 50 15 15


Route Advertising

Distance Vector style protocolDistance Vector style protocol

Hear advertisements: IP prefix, AS-pathHear advertisements: IP prefix, AS-path

Filter if desired (i.e. ignore)Filter if desired (i.e. ignore)

Append yourself: IP prefix, myAS+AS-pathAppend yourself: IP prefix, myAS+AS-path

Forward to appropriate ASs Forward to appropriate ASs


Basic AS relationships

Customer – ProviderCustomer – Provider• Customer pays Provider for serviceCustomer pays Provider for service• The Customer is always rightThe Customer is always right

Peer to Peer: mutual cooperationPeer to Peer: mutual cooperation• Ex. MCI and AT&TEx. MCI and AT&T

Sibling-SiblingSibling-Sibling• Ex. AT&T research and AT&T wirelessEx. AT&T research and AT&T wireless


The Internet as a Directed Graph

Every edge is Every edge is bidirectionalbidirectional

Business relationships Business relationships are representedare represented

Provider Customer

Peer Peer


The Initial Idea

Data flows between customers-providersData flows between customers-providers

Top level providers are peersTop level providers are peers• They exchange information to ensure connectivityThey exchange information to ensure connectivity

What can possibly go wrong?What can possibly go wrong?


And then came the rain…

Thousands of ASsThousands of ASs

Complicated relationshipsComplicated relationships

Multiple providers for one AS!!Multiple providers for one AS!!• MultihomingMultihoming

Traffic engineeringTraffic engineering• I want to use multiple paths and load balanceI want to use multiple paths and load balance


Example: The Intended Use

Provider Customer

Peer Peer


BGP Graph and Routing Policies

Up then down: 1, 10, 100, 200, 13, 4Up then down: 1, 10, 100, 200, 13, 4No valleys, no up-down-up, no more than 1 peer-peer No valleys, no up-down-up, no more than 1 peer-peer

10

1

100 200

11

4

12 13

2 3


The Rules of BGP Routing

Transit traffic: traffic that does not go to my Transit traffic: traffic that does not go to my customers (or their customers)customers (or their customers)

A provider carries any traffic to, from A provider carries any traffic to, from customercustomer

Peers exchange traffic Peers exchange traffic onlyonly if between their if between their customerscustomers


Implementing BGP Rules

A customer advertises whatever she wants A customer advertises whatever she wants

A provider forwards everything from customerA provider forwards everything from customer• So that world now where customer IPs areSo that world now where customer IPs are

A provider advertises whatever customer wants to A provider advertises whatever customer wants to customercustomer

A peer hears but does not advertise further prefixes A peer hears but does not advertise further prefixes from a peerfrom a peer

A peer advertises only its customers to a peerA peer advertises only its customers to a peer


Some Simple Policies: Transitivity

• A customer should not be transit for its ProvidersA customer should not be transit for its Providers• For this: ASX should not advertise ISP1 advertisementsFor this: ASX should not advertise ISP1 advertisements

ISP 1 ISP 2

AS X

Not allowed


Basic AS Relationships

Customer – Provider: Customer – Provider: customer pays and is always rightcustomer pays and is always right

Peer to Peer: Peer to Peer: Exchange traffic only between their customersExchange traffic only between their customers

Sibling-Sibling: Sibling-Sibling: Exchange traffic at willExchange traffic at will

10

1

100 200

11

4

12 13

2 3

Provider Customer

PeerPeer


How BGP Policy Restricts Routing

Routing rules:Routing rules:• Provider accept everythingProvider accept everything• Peer only if it is for its Peer only if it is for its

customerscustomers

Path Properties:Path Properties:• Up then downUp then down• No up-down-up, at most 1 No up-down-up, at most 1

peer-peer stepspeer-peer steps

Provider Customer

10

1

100 200

11

4

12 13

2 3

Peer Peer


How Policies Affect Routing

A Provider will get rid of A Provider will get rid of traffic as soon as traffic as soon as possible,possible,

But a Provider will carry But a Provider will carry the traffic for its the traffic for its customercustomer

Did anyone say traffic is Did anyone say traffic is asymmetric?asymmetric?

Customer 1

ISP1

ISP2Customer 2


BGP Path-Length Asymmetry

Consider number of AS traversed by a pathConsider number of AS traversed by a pathAsymmetry: 46% of pairs differ by at least one AS hopAsymmetry: 46% of pairs differ by at least one AS hop

[Siganos 01][Siganos 01]


Determining The Logical Graph

The business relationships are criticalThe business relationships are critical

How can I find the relationships?How can I find the relationships?1.1. Infer relationships from routing tablesInfer relationships from routing tables

2.2. IRR database: manually maintained – error proneIRR database: manually maintained – error prone


Two Inference Algorithms

Inference algorithm Inference algorithm [Gao 00][Gao 00]

• Using 1 routing tableUsing 1 routing table• Exploit the up-down path propertyExploit the up-down path property

in a routing path, assume highest degree node as peakin a routing path, assume highest degree node as peak

Inference using multiple tables Inference using multiple tables [Subramanian02][Subramanian02]

• Use multiple points of observation to improve resultsUse multiple points of observation to improve results


Things Become Interesting:Traffic Engineering

How can I pick a route?How can I pick a route?Local Preference: path Local Preference: path attributeattributeAS2 wants to prefer fast AS2 wants to prefer fast thick linkthick linkAdvertisement from right Advertisement from right router of AS2 has higher router of AS2 has higher Local PreferenceLocal PreferenceAny BGP router in AS2 will Any BGP router in AS2 will prefer the prefer the

AS 1

AS 2

208.1.1.0/24

LP 100LP 80

slow


Load Balancing - Appetizer

I want to share traffic between my two I want to share traffic between my two providersproviders

How can I do this?How can I do this?


Load Balancing: Long Prefix Match Takes All!

Customer

ISP 1 ISP 2

ISP 3

138.39.1/24

138.39/16138.39.1/24

138.39.1/24

138.39/16

138.39.1/24


So How Can I Balance the Load?

Ask my provider to not aggregate my prefixAsk my provider to not aggregate my prefix• Will this work?Will this work?

Split my prefix in twoSplit my prefix in two• 138.39.1.0/24138.39.1.0/24• A: 138.39.1.31/28A: 138.39.1.31/28

• B: B: 138.39.1.32/28138.39.1.32/28

Advertise only one part to ISP2Advertise only one part to ISP2ISP2 traffic destined for prefixes in AISP2 traffic destined for prefixes in AISP1 traffic destined for prefixes in BISP1 traffic destined for prefixes in B


Summary Up to Now

BGP-4 is the de facto protocol for BGP-4 is the de facto protocol for interdomain routinginterdomain routing

BGP was developed to achieve:BGP was developed to achieve:• Flexible policy implementationFlexible policy implementation• Scalability via route aggregation given CIDRScalability via route aggregation given CIDR

There are many open issuesThere are many open issues• BGP is a hot research topicBGP is a hot research topic


The Growth of BGP Table


The BGP Growth: The Truth

Growth flattened Growth flattened outout

Why?Why?• Better Better

managementmanagement• Dot-com crash?Dot-com crash?


Routing Table Variation

Larger ASes have significantly larger tablesLarger ASes have significantly larger tables


Update Activity Per Prefix

Measure rate of announcements + Measure rate of announcements + withdrawals + path updateswithdrawals + path updates

Compare relative update rate per prefixCompare relative update rate per prefix

length to the relative number of length to the relative number of prefixes of that lengthprefixes of that length>1 implies higher than average update >1 implies higher than average update

rate (less stable)rate (less stable)<1 implies lower than average update <1 implies lower than average update

rate(more stable)rate(more stable)


Measured Update Rate

Bursty!Bursty!


BGP robustness

Measuring the BGP updatesMeasuring the BGP updates


Analyzing Messages By Content

Aggregated per 30 secondsAggregated per 30 seconds


Initial Observations

Updates show daily and weekly periodicityUpdates show daily and weekly periodicityThere is no evidence BGP disturbance:There is no evidence BGP disturbance:• The Baltimore tunnel train 18 July that destroyed The Baltimore tunnel train 18 July that destroyed

Internet linesInternet lines• Sept 11 attackSept 11 attack

There are some spikes at:There are some spikes at:• 19 July19 July• 18-22 September18-22 September


BGP Updates Correlations


BGP Under Attack


Router CPU Activity Correlates…


The Attack of The Worm


Classification of Instabilities


Isolating Instability: 1 unstable peer


Detecting abnormal BGP activity


The Worm Activity


The Worm Correlates Again…


Summary of BGP Instability

Globally correlated BGP instability is not Globally correlated BGP instability is not uncommonuncommon

Some causes are well understood Some causes are well understood (misconfiguration, bad path announcements)(misconfiguration, bad path announcements)

Some others are less well understood, and Some others are less well understood, and more worrisome:more worrisome:• wormsworms


BGP: Up Close and Personal

Establishing a connectionEstablishing a connection

MessagesMessages

Path AttributesPath Attributes


Establishing A BGP Session

BGP uses TCP connectionsBGP uses TCP connections• For reliabilityFor reliability

A BGP session is between two routersA BGP session is between two routers• Typically directly connected (Ethernet, FDDI)Typically directly connected (Ethernet, FDDI)

Routers establish a BGP sessionRouters establish a BGP session• Authentication and set-upAuthentication and set-up• Update and withdrawalsUpdate and withdrawals• If disconnected, all paths are invalidatedIf disconnected, all paths are invalidated


Messages

First open TCP connectionFirst open TCP connection• Identification and authenticationIdentification and authentication

BGP messagesBGP messages• OPEN: set-up, negotiate timer for keep-aliveOPEN: set-up, negotiate timer for keep-alive• UPDATE: routing changesUPDATE: routing changes• NOTIFICATION: termination, and error messagesNOTIFICATION: termination, and error messages• KEEPALIVE: confirm that connection is active KEEPALIVE: confirm that connection is active


UPDATE Message

Advertise reachability informationAdvertise reachability information

Withdraw paths to prefixesWithdraw paths to prefixes

Update information prefixesUpdate information prefixes

Introduce new prefixesIntroduce new prefixes

Modify important path attributes for new Modify important path attributes for new prefixes and the related pathsprefixes and the related paths


Standard Path Attributes

Attribute: Type, Length, ValueAttribute: Type, Length, Value

Origin: where did I hear this fromOrigin: where did I hear this from• BGP (external) or IGP (internal)BGP (external) or IGP (internal)

AS Path: sequence of ASsAS Path: sequence of ASs

Flexible handling of loopsFlexible handling of loops• Recovering from disconnected ASs!Recovering from disconnected ASs!

Next Hop: set explicitly who the next router Next Hop: set explicitly who the next router should be (possibly a non BGP speaker)should be (possibly a non BGP speaker)


Multi-Exit Discriminator (MED)

•Indication to external peers of the preferred path into the AS•Lowest Med Preferred

MED 10MED 10

MED 50MED 50

AS 1

AS 2


Note for MED

One AS sets the valuesOne AS sets the values

Another AS interprets and uses themAnother AS interprets and uses them

Thus:Thus:• Cooperative ASesCooperative ASes• Only between two ASes (1 hop scope)Only between two ASes (1 hop scope)

MED is meaningless in the next hopMED is meaningless in the next hop

MED can be used only if both routes are MED can be used only if both routes are advertised from the same ASadvertised from the same AS


Local Preference

Choosing paths Choosing paths internally internally Within an AS:Within an AS:• Set Local Set Local

Preference to pick Preference to pick the path you want the path you want to send data toto send data to

The higher Local The higher Local Preference is Preference is preferredpreferred

AS 1AS 2

L2

L1

AS3

Here MED can not be used since we have different AS


Aggregation of updates

AS want to aggregate as much as possibleAS want to aggregate as much as possible• Reduce routing stateReduce routing state• Reduce information that needs to be exchangedReduce information that needs to be exchanged

Main idea: send one update instead of twoMain idea: send one update instead of two• All other attributes are the same (path, preferences etc)All other attributes are the same (path, preferences etc)• The prefixes have to be subsets, or adjacentThe prefixes have to be subsets, or adjacent

subset adjacent


Atomic Aggregate

Sometimes we aggregate paths that are Sometimes we aggregate paths that are different (ie different AS sequence)different (ie different AS sequence)

Atomic aggregate shows that some of the Atomic aggregate shows that some of the destinations in this prefix are not necessarily destinations in this prefix are not necessarily following the same pathfollowing the same path

This does not allow other routers to de-This does not allow other routers to de-aggregate the path, thus creating entries that aggregate the path, thus creating entries that should not appearshould not appear


Atomic Aggregate

AS 2 thinks that all 138.39/16 follow the 3 1 4 pathAS 2 thinks that all 138.39/16 follow the 3 1 4 pathAS 5

AS 1AS 2

AS 3

138.39.1/24

138.39/16138.39.1/24

138.39/16 : 1 4

AS 4 138.39/16

138.39/16 : 3 1 4


BGP Route Selection Process

1.1. Maximum prefix length matchMaximum prefix length match2.2. Highest Local PriorityHighest Local Priority3.3. Shortest AS PathShortest AS Path4.4. Lowest MED (if routes through same AS)Lowest MED (if routes through same AS)5.5. Min Cost Next hop router (consulting IGP)Min Cost Next hop router (consulting IGP)6.6. Prefer external to internal routesPrefer external to internal routes

1.1. Pick lowest BGP identifier among many E-BGPPick lowest BGP identifier among many E-BGP2.2. Pick lowest BGP identifier among many I-BGPPick lowest BGP identifier among many I-BGP


Internal BGP (I-BGP)

Communication between routers of an ASCommunication between routers of an ASI-BGP very similar to E-BGP except:I-BGP very similar to E-BGP except:• Different advertising rulesDifferent advertising rules

Do not re-advertise a path internallyDo not re-advertise a path internally• I-BGP 1 -> I-BGP 2 -X- I-BGP3I-BGP 1 -> I-BGP 2 -X- I-BGP3

But readvertise (in -out, out -in):But readvertise (in -out, out -in):• I-BGP 1 -> I-BGP 2 -> E-BGP1I-BGP 1 -> I-BGP 2 -> E-BGP1• E-BGP 1 -> I-BGP 2 -> I-BGP1E-BGP 1 -> I-BGP 2 -> I-BGP1


I-BGP Re-Advertising

B will not re-ad. to C B will not re-ad. to C what it hears from Awhat it hears from ABut it will re-ad to DBut it will re-ad to DWhy?Why?Paths are identified Paths are identified by AS, and internally by AS, and internally you have the same you have the same ASAS• To avoid routing loopsTo avoid routing loops

E-BGPI-BGP

A B

C

D


I-BGP Mesh

I-BGP routers form a fully connected meshI-BGP routers form a fully connected mesh

(clique)(clique)• Scalability becomes an issueScalability becomes an issue

The full mesh is independent of physical The full mesh is independent of physical connectivityconnectivity


A Subtle Difference I-BGP vs E-BGP

E-BGP: routers communicate using the IP of E-BGP: routers communicate using the IP of the physical interface (link)the physical interface (link)• Link based reliabilityLink based reliability

I-BGP: routers have “virtual or loopback” I-BGP: routers have “virtual or loopback” interfaceinterface• Even when link fails, routers may be reachableEven when link fails, routers may be reachable• Node based reliabiityNode based reliabiity


Static vs Dynamic Configuration

In practice, many customers do not speak In practice, many customers do not speak BGPBGP

Do not have an AS numberDo not have an AS number

They are configured staticallyThey are configured statically


One vs Many Providers

Single homed vs multihomedSingle homed vs multihomed

Customers for reliability and performance Customers for reliability and performance connect to many providersconnect to many providers

Difference:Difference:• Single homed: easy to manageSingle homed: easy to manage• Multihomed: trickyMultihomed: tricky

Route aggregationRoute aggregation Load balancingLoad balancing


Multihoming

How can I use my multiple connections How can I use my multiple connections efficiently?efficiently?

Multihoming is quite widespreadMultihoming is quite widespread

Users take it up to themselves [Huston]Users take it up to themselves [Huston]

Consequence: non-aggregatable stateConsequence: non-aggregatable state


Two routers two links

Multiplex traffic at link or IP layerMultiplex traffic at link or IP layer

Use Next Hop to point to virtual router IPUse Next Hop to point to virtual router IP• Second look up: how can I reach virtual IP?Second look up: how can I reach virtual IP?• Pick one of the two links randomly or staticallyPick one of the two links randomly or statically• Furthermore, when one link is down the other one is Furthermore, when one link is down the other one is

chosenchosen


Multihoming with One Routers 1 ISP

Customer can advertise Customer can advertise different prefixes on each different prefixes on each linklink• Reliability?Reliability?

Use Multi-Exit-DiscriminatorUse Multi-Exit-Discriminator

Use Local PreferenceUse Local Preference• ISP to pick linkISP to pick link• Customer’s IBGP to pick routerCustomer’s IBGP to pick router

C A

ISP 1

P1 P2

B


Issues

The mechanisms are there, but they depend The mechanisms are there, but they depend on many factorson many factors• How much traffic each prefix generates?How much traffic each prefix generates?

Balancing incoming and outgoing trafficBalancing incoming and outgoing trafficDynamically adapting to changing conditionsDynamically adapting to changing conditionsTechnical issues:Technical issues:• Some ISPs do not accept very long prefixesSome ISPs do not accept very long prefixes

Longer than they would be in classful routingLonger than they would be in classful routing Longer than 19 for new prefixesLonger than 19 for new prefixes


Multihoming with Two Routers 1 ISP

Similar case for Provider to Similar case for Provider to CustomerCustomerFor Customer to Provider:For Customer to Provider:• A could alternate pathsA could alternate paths

Reordering of packetsReordering of packets• ISP 1 could advertise different ISP 1 could advertise different

addresses on each linkaddresses on each link I.e. 1: ISP customersI.e. 1: ISP customers 2: Default (everybody 2: Default (everybody

else)else)

BC

A

ISP 1

P1 P2

Internet

ISP 1Customers


Multihoming With Two Providers

Customer

ISP 1 ISP 2

ISP 3

138.39.1/24

138.39/16138.39.1/24

138.39.1/24

138.39/16

138.39.1/24


Two Providers Multihoming: Getting Address Space

Given two ISPGiven two ISP

Get space from one (we saw before)Get space from one (we saw before)

Get space from bothGet space from both• Advertise only related prefix to ISPAdvertise only related prefix to ISP

Aggregation but not reliabilityAggregation but not reliability• Advertise both prefixes to ISPsAdvertise both prefixes to ISPs

No aggregation but reliabilityNo aggregation but reliability

Get space independently of bothGet space independently of both• Max flexibility, not reliabilityMax flexibility, not reliability• If too narrow of prefix, may not propagate, no connectivityIf too narrow of prefix, may not propagate, no connectivity


I-BGP Scalability

Full mesh is not scalable: Full mesh is not scalable: O(N^2) IBGP sessionsO(N^2) IBGP sessions

Approaches to scalable I-BGPApproaches to scalable I-BGP• Hierarchical structure: Route reflectorsHierarchical structure: Route reflectors• Divide and conquer: ConfederationsDivide and conquer: Confederations


Route Reflection

Explicitly allow some I-Explicitly allow some I-BGP routers to BGP routers to readvertisereadvertise

Route reflectors: Route reflectors: represent other routersrepresent other routers

Hierarchical structure Hierarchical structure avoids loops and avoids loops and problemsproblems

B

AC

D E


Confederations

Decompose an AS to Decompose an AS to sub-ASsub-AS

Externally: one ASExternally: one AS

Internally: like E-BGP Internally: like E-BGP (E-I-BGP)(E-I-BGP)

Loop avoidance:Loop avoidance:• AS-CONFED-SETAS-CONFED-SET• AS-CONFED-SEQUENCEAS-CONFED-SEQUENCE

B

AC

D E


Confederation BGP Rules

Differences of BGP between sub-ASesDifferences of BGP between sub-ASes

Local-Preference is transitiveLocal-Preference is transitive• Recall in BGP it is notRecall in BGP it is not

NEXT-HOP attribute is forwardedNEXT-HOP attribute is forwarded

Path within AS is monitored with sub-AS no.Path within AS is monitored with sub-AS no.• AS-CONFED-SEQAS-CONFED-SEQ• AS-CONFED-SEQUENCEAS-CONFED-SEQUENCE


Handling the Dynamic Nature


To Refresh or not to Refresh?

BGP uses hard state:BGP uses hard state:

BGP routers consider a path is “usable” BGP routers consider a path is “usable” • until explicitly withdrawnuntil explicitly withdrawn• the session failsthe session fails

How do I detect if a connection failed?How do I detect if a connection failed?• Keep-alive messagesKeep-alive messages


Path Updates Frequency

Send updates of a path no sooner 30 secSend updates of a path no sooner 30 secWhy?Why?• Stability Stability • Overhead reductionOverhead reduction

Side-effectsSide-effects• Convergence can take longerConvergence can take longer

What is the right interval?What is the right interval?• Recent studies say that 30s is too longRecent studies say that 30s is too long


Route Flapping and Damping

Flapping: constant alternating updatesFlapping: constant alternating updates

It can happen!It can happen!

Route update dampening:Route update dampening:• Consider stability of path before using itConsider stability of path before using it

How store “penalty” value for each pathHow store “penalty” value for each path

Issue: router needs to remember withdrawn Issue: router needs to remember withdrawn pathspaths


Overview up to Now

BGP is Distance VectorBGP is Distance Vector

BGP uses TCP and hard-stateBGP uses TCP and hard-state

Routing updates are “delayed” and batchedRouting updates are “delayed” and batched

Route dampening to alleviate instabilitiesRoute dampening to alleviate instabilities


End


Current Research: the AS Topology

Characterize the topology using power-laws Characterize the topology using power-laws (Faloutsos 99, Siganos et al 01)(Faloutsos 99, Siganos et al 01)

Modeling the evolution of the topologyModeling the evolution of the topology (Barabasi, Siganos01)(Barabasi, Siganos01)


Ongoing Research: AS paths

107 Gb of AS paths over 3 years107 Gb of AS paths over 3 years

Exponential growth, but distances remain the Exponential growth, but distances remain the same same

Inflation due to policy:Inflation due to policy:• 20% of paths are larger than they could20% of paths are larger than they could

Significant Routing Asymmetry:Significant Routing Asymmetry:• 40% paths by at least one hop40% paths by at least one hop


Near Future Research Plans

Mine the collected paths for patternsMine the collected paths for patterns

How stable were the paths?How stable were the paths?

Can we identify “illegal” paths?Can we identify “illegal” paths?

Identify pathologies (ie. loops) Identify pathologies (ie. loops)


Overview of Plans for Research

How well does BGP work now?How well does BGP work now?

How secure and robust is BGP?How secure and robust is BGP?• Cyber-terrorism: how much damage can be done?Cyber-terrorism: how much damage can be done?

How would we design BGP now?How would we design BGP now?• People are asking this questionPeople are asking this question

How well will BGP scale?How well will BGP scale?

How can we manage BGP (avoid human errors)?How can we manage BGP (avoid human errors)?

Approach:Approach:• Analytical and simulations with SSFNETAnalytical and simulations with SSFNET


Conclusions

BGP is an open and exciting topicBGP is an open and exciting topic

The community knows very littleThe community knows very little

Big ticket items:Big ticket items:• Measurments and modelingMeasurments and modeling• Robustness, security Robustness, security • Network Management: traffic engineeringNetwork Management: traffic engineering• ScalabilityScalability


Multihoming With Two Providers

Customer

ISP 1 ISP 2

ISP 3

138.39.1/24

138.39/16138.39.1/24

138.39.1/24

138.39/16

138.39.1/24


Policies on Transitivity

NontransitA customer should notBe transit for its Providers

Isp 1 Isp 2

AS X

Not allowed

TransitAS 1

AS 2

AS 3 AS 4


More BGP attributes

•Communities•Confederations•Route Reflectors


Convergence


Multi-Exit Discriminator(MED)

•Indication to external peers of the preferred path into the AS•Lowest Med Preferred


Note in MED

One AS sets the valuesOne AS sets the values

Another AS interpets and uses themAnother AS interpets and uses them

Thus:Thus:• Cooperative ASesCooperative ASes• Only between two ASes (1 hop scope)Only between two ASes (1 hop scope)

MED is meaningless in the next hopMED is meaningless in the next hop

MED can be used only if both routes are MED can be used only if both routes are advertised from the same ASadvertised from the same AS


Local Preference

Choosing paths internally Choosing paths internally

Within an AS:Within an AS:• Set Local Preference to pick the path you want to Set Local Preference to pick the path you want to

send data tosend data to

The higher Local Preference is preferredThe higher Local Preference is preferred


Atomic Aggregate

Sometimes we aggregate paths that are Sometimes we aggregate paths that are different (ie different AS sequence)different (ie different AS sequence)

Atomic aggregate shows that some of the Atomic aggregate shows that some of the destinations in this prefix are not necessarily destinations in this prefix are not necessarily following the same pathfollowing the same path


Aggregator

Notify that an aggregation took placeNotify that an aggregation took place• Which ASWhich AS• Which routerWhich router

For management and traceback purposesFor management and traceback purposes

BGP: Introduction and Issues

Documents

Transcript of BGP: Introduction and Issues