ROUTING, PART 1reeves.csc.ncsu.edu/Classes/csc573/routing1.pdf · Scalability – + Convergence Time – + Property DV LS copyright2005DouglasS.Reeves 34 Also… •Flooding and broadcasting

1

ROUTING, PART 1

Internet Protocols

CSC / ECE 573

Fall, 2005

N. C. State University

copyright 2005 Douglas S. Reeves 2

Today’s Lecture

I. Basic concepts and terminology

II. Distance-Vector protocols

III. Link-State protocols


Announcements

Exam will be returned on Tuesday

HW4 online – due Tuesday

help!

Project Part II online

BASIC CONCEPTS ANDTERMINOLOGY


Routing Protocols

• Goal: configure router forwarding tables sodatagrams from anywhere to anywhere will bedelivered by the “best” path– what is “best”?

– what info does each router need?– how does the routing algorithm work?

– how ensure routers make consistent decisions?

• Steps1. routers discover their neighbors

2. routers exchange information about the network

3. routers compute forwarding tablescopyright 2005 Douglas S. Reeves 6

Metrics For Routing Algorithms1. Find a route if one exists

2. Avoid routing loops

3. Support for enforcement of routing “policy”

4. Scale to large size networks

5. Adapt quickly to changes in network state

6. Recover quickly from temporary failures andmistakes

7. Minimize size of routing tables

2


What is the “Best” Route?


Autonomous Systems (RFC 1930)

• An autonomous system (AS) is an independentrouting domain within which all routers agree onthe same external routing policy– under control of a single administrative entity

– each AS gets a 16 bit ID number– there are currently ~17,000 ASes

• AS’es are arranged in a hierarchy; those at top…– have largest routing tables– have most specific routes

– do not use a “default” route


Internet Routing Architecture

• Regional registries (ARIN, RIPE, LACNIC, etc.)– manage allocation of CIDR blocks– manage allocation of AS numbers

– manage allocation of top-level domain names

• Interconnection of ISPs occurs by peering ortraffic exchange agreement– either they exchange routing tables, and carry traffic

for each other, or

– one network serves as transit network for another


Internet Routing Architecture (cont’d)

• Facilities where ISPs interconnect andexchange traffic are Internet Exchanges– slightly less than 100 in the world today

• A route server collects route information from allpeers at an exchange, and redistributes to them

• The Internet Routing Registry is a distributeddatabase where operators register their routinginformation– allows consistency checking of routing policies


Autonomous Systems Example

Default-freeProvider

Default-freeProvider

Mid-levelProvider

Mid-levelProvider

Dial-upProvider

Dial-upProvider

Company Company

RouteServer

RouteServer

InternetExchange

InternetExchange


The Internet AS Topology, Again

3


Distribution of AS Degree (2003)Degree Histogram of ASes on AS-level Internet Topology

(12538 ASes, 32932 Links / Skitter 8/2003)

1

10

100

1000

10000

1 10 100 1000

Number of Peers of the Autonomous System

Fre

qu

en

cy


Interior vs. Exterior Gateway Protocols

• Interior (intra-domain): within a single AS– single administration, single policy– normal objectives: efficiency / simplicity / stability

• Exterior (inter-domain): between ASes– many organizations, each with own policy– objectives: configurability, scalability, robustness,

security

• Exchange of info between inter-domain, intra-domain routing occurs at border gateways

DISTANCE VECTOR PROTOCOLS


Distance Vector (DV) Routing

• Routers exchange information in the followingform:<destination1, min_cost to destination1>,<destination2, min_cost to destination2>,<destination3, min_cost to destination3>, ...

• Routers do not know the complete networktopology

• Exchange info with directly-attached neighborsonly– multiple steps of exchange needed to propagate

information across the network


DV Route Computation• Asynchronous, iterative, distributed computation

• For each destination, find…– cost of least cost path– “first hop” (neighbor) router on least cost path

• Computation based on Bellman-Ford algorithm


DV Route Computation (cont’d)

• In words:Least cost path to destination j is the path startingfrom any neighbor k for which(cost of link to k + path cost from k to j)is the least

4


DV Routing Example


DV Routing Example (cont'd)


DV Routing Example (cont'd)


The Spanning Tree from A


Results of DV Computation

• Guaranteed non-looping LC path between everypair of nodes, after convergence

• But, while converging, or when changes occur,there may be temporary routing loops orinefficient paths

• Source of the problem: with distance vectors, youcan’t tell (locally) what path is being taken to adestination

DV Routing Example -- Link Failure

i.e., D’s distance vector has not been updated (yet),so D’s LC path to B goes through A,and A’s LC path to B goes through D

5


Propagating Distance Information

Distance tonetwork N1

N1 N2 N3 N4 N5

R1 R2 R3 R4 R5

InitiallyInfInfInfInf

After 1 exchange (DV’s)InfInfInf1

After 2 exchangesInfInf21

After 3 exchangesInf321

After 4 exchanges4321

• “Good news travels fast” = availability of a newlink or router


The “Count to Infinity” Problem


N1 N2 N3 N4 N5

R1 R2 R3 R4 R5

• “Bad news travels slowly” = failure of an existinglink or router

Initially4321

link fails


The “Count to Infinity” Problem


N1 N2 N3 N4 N5

R1 R2 R3 R4 R5




• “Bad news travels slowly” = failure of an existinglink or router

Initially4321



link fails

LINK-STATE PROTOCOLS


“Link State” (LS) Routing

• A link-state database is…

– a map of the entire network topology, describing nodesand links

• Each router maintains an identical copy of thisdatabase– routers broadcast link-state advertisements (LSAs) to

update the database of other routers– each router is responsible for originating an LSA

describing its working interfaces


LS Database Example

6


Computing Routes from LSAs

• Each router computes the least cost path fromitself to all networks– normally computed by Dijkstra's shortest-path

algorithm– no convergence issues, no “count to infinity” issues

– only issue: how long to propagate LSAs throughoutentire network?

• With complete network topology information,there is more flexibility about how to optimizeroutes


Example: LSAs


Comparison of LS and DV Algorithms

–+Amount of data exchanged–+Simplicity

??Robustness+–Flexibility (metrics)+–Scalability+–Convergence Time

LSDVProperty


Also…

• Flooding and broadcasting

• Hierarchical routing


Summary

The Internet routing architecture consists ofrouters, Internet exchanges, routing databases

The top level of the Internet organization is theAutonomous Systems level

Exterior gateway protocols exchange routinginformation between ASes; interior protocols areused inside an AS


Summary (cont’d)

The two classes of routing algorithms aredistance-vector and link-statelink-state is more powerful, converges faster, and

scales betterdistance vector is simpler

7


Next Lecture

• Routing, Part 2 (RIP and OSPF)

ROUTING, PART 1reeves.csc.ncsu.edu/Classes/csc573/routing1.pdf · Scalability – + Convergence Time – + Property DV LS copyright2005DouglasS.Reeves 34 Also… •Flooding and broadcasting

Documents