Network Routing:

Link State Routing and BGP

4/2/2008

Admin.

Course feedback two more written assignments one programming assignment (not assigned

yet, but already linked on the schedule page just in case you want to play with it; read Chapter 7)

Assignment 3

Recap: Distance Vector Routing

Implemented in RIP, IGRP, EIGRPMany properties

Distributed Asynchronous Convergence in finite steps,

independent of initial condition if network is connected

Good news propagate fast

Counting-to-infinity problem

• reverse poison/split horizon

Outline

Recap Distance vector protocols

o synchronous Bellman-Ford (SBF)o asynchronous Bellman-Ford (ABF) destination-sequenced distance vector

(DSDV)

Destination-Sequenced Distance Vector protocol (DSDV)

An extension of distance vector protocol to address the counting-to-infinity problem

Extension DSDV tags each route with a sequence number each destination node D periodically advertises

monotonically increasing even-numbered sequence numbers

when a node realizes that its link to a destination is broken, it advertises the route to D with an infinite metric and a sequence number which is one greater than the previous route (i.e. an odd seq. number)

• the route is repaired by a later even-number advertisement from the destination

DSDV: More Detail

Let’s assume the destination node is D

There are optimizations but we present a simple version: each node maintains only (SB, dB), where SB is the

sequence number at B for destination D and dB is the best distance using a neighbor from B to D

Both periodical and triggered updates periodically: D increases its seq. by 2 and broadcasts

with (SD, 0) if B is using C as next hop to D and B discovers that C

is no longer reachable• B increases its sequence number SB by 1, sets dB to , and

sends (SB, dB) to all neighbors

A B

route update

DSDV: Update

Update after receiving a message assume B sends to A its current state (SB, dB) when A receives (SB, dB)

– if SB > SA, then // always update if a higher seq#

» SA = SB

» if (dB == ) dA = ; else dA= dB + d(A,B)

– else if SA == SB, then

» if dA > dB + d(A,B) // update for the same seq# only if better route dA= dB + d(A,B) and uses B as next hop

A B

route update

Example

When C discovers that C-D link is down, it increases its seq# and broadcasts its cost to be

1

1 1

1

Claim: DSDV Does Not Form Loop

what is a loop? a loop is a global state (consisting of the nodes’ local

states) at a global moment (observed by an oracle) such that there exist nodes A, B, C, … E such that A (locally) thinks B as down stream, B thinks C as down stream, … E thinks A as down stream

Initially no loop (no one has next hop so no loop) Derive contradiction if a loop forms after a node

processes an update, e.g., when A receives the

update from B, A decides to use B as next hop and forms a loop

A B

update

Background: Global Invariants

This is a very effective method in understanding distributed asynchronous protocols

Invariants are defined over the states of the distributed nodes

Consider any node B. Let’s identify some invariants over the

state of node B, i.e., (SB, dB).

Invariants of a Single Node B

Some invariants about the state of a node SB is non-decreasing

dB is non-increasing for the same sequence number

time

Invariants of if A Considers B as Next Hop

Some invariants if A considers B as next hop

SA cannot be an odd number, dA is not

SB SA

because A is having the seq# which B last sent to A; B’s seq# might be increased after B sent its state

• if SB == SA then dB < dA because dA is based on dB which B sent to A some time ago, dB < dA since all link costs are positive; dB might be decreased after B sent its state

A B

Loop Freedom of DSDV

Consider a critical moment A starts to consider B as

next hop, and we have a loop If any link in the loop

(X considers Y as next hop) satisfies SY > SX

• by transition along the loop SB > SB

If all nodes along the loop have the same sequence number

• by transition along the loop dB > dB

A B

X

Y

update

Summary: DSDV

DSDV uses sequence number to avoid routing loops seq# partitions routing updates from different

outside events within same event, no loop so long each node

only decreases its distance

EIGRP: a routing protocol by Cisco routers Diffusive Update Algorithm (DUAL)

Discussion: Distance Vector Routing

What do you like about distance vector routing?

What do you not like about distance vector routing?

Churns of DV: One Example

N+2

…

…1 2 3 N-1 N

N+1

N+3

2N

Initial Conditions + All links have cost 1

Problematic Message sequences1. Node 2 tells 3. Node 3 tells 4…

Node N tells N+1. (N-1 messages)

2. Node N+1 tells N+2, N+2 tells N+3,…,2N. (N-1 messages)

3. Now node N-1 tells node N+14. Step 2 repeats5. Now node N-2 tells node N+16. …

A total of N2 – 2 messages

Question to think about: is this the worst case?

Outline

Recap Distance vector protocols Link state protocols

Link-State Routing

Net topology, link costs are distributed to all nodes all nodes have same info thus can compute any types of routes

Each node computes its shortest paths from itself to all other nodes e.g., use Dijkstra’s algorithm (see backup slides)

Link state distribution accomplished via “link state broadcast”

Link State Broadcast

The hard part is link state broadcast basic approach: forward a link state (link ID,

link status) to all links except the incoming link

Question: what are the problems the link state broadcast needs to handle? broadcast loop (a problem in exam 1) ordering of events (link up and down) network partitioning and then merge

Link State Broadcast

Each link update is given a sequence number: (initiator, seq#, link, status) the initiator should increase the seq# for each

new update If the seq# of an update of a link is not

higher than the highest seq# a router has seen, drop the update

Otherwise, forward it to all links except the incoming link (see backup slides for the real implementation using packet buffer)

Each seq# has an age field (why?) Updates are sent periodically (why?)

OSPF (Open Shortest Path First)

“ Open”: publicly available

Uses Link State algorithm link state (LS) packet dissemination topology map at each node route computation using Dijkstra’s algorithm

http://en.wikipedia.org/wiki/Open_Shortest_Path_First

OSPF “Advanced” Features (not in RIP)

Multiple same-cost paths allowed (only one path in RIP)

For each link, multiple cost metrics for different Type Of Service (eg, satellite link cost set “low” for best effort; high for real time)

Security: all OSPF messages authenticated (to prevent malicious intrusion); TCP connections used

Hierarchical OSPF

Hierarchical OSPF

“ summarize” distances to nets in own area, advertise to other Area Border routers.

run OSPF routing limited to backbone.

- Link-state advertisements only in area each nodes has detailed area topology;- only know direction (shortest path) to nets in other areas.

Two-level hierarchy: local area, backbone.

Information hiding (filtered) => reduce computation, bandwidth, storage

Why Hierarchy?

Discussion: Link State Routing

What do you like about link state routing?

What do you not like about link state routing?

Question to think about: which routing protocol (DV or LS) should the Internet use?

Outline

Recap Distance vector protocols Link state protocols Routing in the Internet

overview

Routing in the Internet

The Global Internet consists of Autonomous Systems (AS) interconnected with each otherAn AS is identified by an AS Number

(ASN), e.g. Yale ASN is 29 try %whois -h whois.arin.net “a

Yale“

Routing with AS Intra-AS

Different AS’s may run different routing protocols A protocol running insides an AS is called an Interior

Gateway Protocol (IGP)• RIP: Routing Information Protocol• OSPF: Open Shortest Path First• IS-IS: very similar to OSPF (or should we say OSPF is very

similar to IS-IS?)• E/IGRP: Interior Gateway Routing Protocol (Cisco)

Inter-AS a protocol runs among AS’s is also called an Exterior

Gateway Protocol (EGP) for global connectivity, a single interdomain routing

protocol

eBGP

b a

AS C(RIP intra routing)

Routing in the Internet: Example

b

AS B(OSPF intra routing)

AS A(OSPF intra routing)

iBGP

inter-AS routing

between A and B

Intra-AS and Inter-AS Routing

Host h2

a

b

b

aaC

A

Bd c

A.a

A.c

C.bB.a

cb

Hosth1

Intra-AS routingwithin AS A

intra-AS routingwithin AS B

border (exterior gateway) routers

interior (gateway) routers

Many Routing Processes on a Single Router

Forwarding Table

OSPFdomain

RIPdomain

BGP

OS kernel

RIP process

RIP routing table

Forwarding Table Manager

OSPF process

OSPF Routing table

BGP process

BGP routing table

Why Partition into Intra- and Inter-AS Routing? This partition allows ASes flexibility to

choose their own intra-AS routing protocols autonomy

By aggregating many destinations inside an AS into a single destination in interdomain routing, it improves scalability the partition is a type of hierarchical routing hierarchical routing improves scalability: only a

small number of routers are involved with outside

Yale Internet Connectivity

Yale

Qwest

default routes 0.0.0.0/0pointing to provider.

132.130.0.0/16128.36.0.0/16

AT&T Internet2

Hierarchical Routing May Pay a Price for Path Quality

AS 4

AS 3

AS 2

AS 1

Outline

Recap Distance vector protocols Link state protocols Routing in the Internet BGP

BGP Setup

Internet Interdomain Routing: BGP BGP (Border Gateway Protocol): the de facto standard Path Vector protocol:

similar to Distance Vector protocol a border gateway sends to a neighbor entire path (i.e.,

a sequence of ASes) to a destination, e.g.,• gateway X sends to neighbor N its path to dest. Z:

path (X,Z) = X,Y1,Y2,Y3,…,Z if N selects path(X, Z) advertised by X, then:

path (N,Z) = N, path (X,Z) XN

Z

BGP Operations (Simplified)

Establish session on TCP port 179

Exchange all active routes

Exchange incremental updates

AS1

AS2

while (connection is ALIVE) exchange UPDATE message select best available route if route changes, export to neigh.

BGP session

BGP Messages

Four types of messages

OPEN: opens TCP connection to peer and authenticates sender

UPDATE: advertises new path (or withdraws old)

KEEPALIVE keeps connection alive in absence of UPDATES; also ACKs OPEN request

NOTIFICATION: reports errors in previous msg; also used to close connection

Why Path Vector?

Path vector prevents counting-to-infinity problem

Path vector allows an AS to define local policies on the ASes of a given path

BGP Routing Decision Process

routing cacheselect best

path

export path to neighbors

route selection

policy: rank paths

export policy: which paths

export to which

neighbors

BGP Route Selection Policy

Typical (Cisco) routeselection policy Highest local pref Shortest AS path length Prefer eBGP over iBGP …

Yale

Qwest

default routes 0.0.0.0/0pointing to provider.

AT&T Internet2

Policy Interactions

2

0

31

2 1 02 0

1 3 01 0

3 2 03 0

4

3

preferred

lesspreferred

The BAD GADGET example:- 0 is the destination - the route selection policy of each AS is to prefer its counter clock-wise neighbor

Policy interaction causes routing instability !

Backup Slides

44

45

Dijkstra’s Algorithm

See schedule page for animation

46

Dijkstra’s Algorithm

Notation: c(i,j): link cost from node i to j; cost infinite if

not direct neighbors D(v): current value of cost of path from source

to destination v p(v): predecessor node along path from

source to destination v N: set of nodes whose least cost path

definitively known

47

Dijkstra’s Algorithm

1 Initialization: 2 N = {A} 3 for all nodes v 4 if v adjacent to A 5 then D(v) = c(A,v) 6 else D(v) = infty 7 8 Loop 9 find w not in N such that D(w) is a minimum 10 add w to N 11 update D(v) for all v adjacent to w and not in N: 12 D(v) = min( D(v), D(w) + c(w,v) ) 13 /* new cost to v is either old cost to v or known 14 shortest path cost to w plus cost from w to v */ 15 until all nodes in N

48

Dijkstra’s Algorithm: Example

Step012345

start NA

ADADE

ADEBADEBC

ADEBCF

D(B),p(B)2,A2,A2,A

D(C),p(C)5,A4,D3,E3,E

D(D),p(D)1,A

D(E),p(E)infinity

2,D

D(F),p(F)infinityinfinity

4,E4,E4,E

A

ED

CB

F

2

2

13

1

1

2

53

5

Use Packet Buffer to Distribute Link States

The right figure shows the packet buffer at router BA, C, F are its neighbors

• Add ACK for reliability• Avoid multiple update packets from the same source at the same router• When a router receives a link update from a source S from neighbor X

• if no state for S or seq# higher than current, add the state for S, set Ack for X, set Send for other neighbors; • if has state for S and seq# of the received update is smaller, ignore; • if received seq# equals to the current seq#, set Ack for X, clear Send to X

•Round-robin among the Ack and Send flags

BGP Route Selection Policy

Typical (Cisco) routeselection policy Highest local pref Shortest AS path length Prefer eBGP over iBGP …

50

Yale

Qwest

default routes 0.0.0.0/0pointing to provider.

AT&T Internet2

51

The Gang of Four

Link State Vectoring

EGP

IGP

BGP

RIP(IS-IS)

OSPF