routing and forwarding - polito

© see page 2Routing - 1

ROUTINGAND FORWARDING

Mario Baldiwww.baldi.info


Copyright NoticeCopyright NoticeThis set of transparencies, hereinafter referred to as slides, isprotected by copyright laws and provisions of InternationalTreaties. The title and copyright regarding the slides (including,but not limited to, each and every image, photography, animation,video, audio, music and text) are property of the authors specifiedon page 1.The slides may be reproduced and used freely by research institutes,schools and Universities for non-profit, institutional purposes. Insuch cases, no authorization is requested.Any total or partial use or reproduction (including, but not limitedto, reproduction on magnetic media, computer networks, and printedreproduction) is forbidden, unless explicitly authorized by theauthors by means of written license.Information included in these slides is deemed as accurate at thedate of publication. Such information is supplied for merelyeducational purposes and may not be used in designing systems,products, networks, etc. In any case, these slides are subject tochanges without any previous notice. The authors do not assume anyresponsibility for the contents of these slides (including, but notlimited to, accuracy, completeness, enforceability, updated-ness ofinformation hereinafter provided).In any case, accordance with information hereinafter included mustnot be declared.In any case, this copyright notice must never be removed and must bereported even in partial uses.


Routing

Determine a paththrough the network

for packets


Advance packets throughthe network

Includes a routingdecision

Forwarding


Forwardingand Routing

A

C

B

D

FEF

B,D

C,D

,FC

,A,B

,D

E,A,B

B,D

A,B

F

A,C,E,F

A,B

C,D

,F

A,C,E,F1. Routing(proactive)

B

B B

2. Forwarding(on‐the‐flyrouting)


Proactive Routing

Independent of actualtraffic

Determine reachabledestinations

Compute best route Commonly referred

to as “routing”


On-the-fly Routing

Realized when handlingeach packet

Based on local informationRouting/forwarding tableOutput of proactiverouting or signaling

A.k.a. route


On-the-fly Routing Algorithms

Routing by Network Address Label Swapping Source Routing

Each protocol architectureadopts one or more


Forwarding Phases

Routing (on-the-fly) Output port selection Possibly next-hopselection

Switching: transferto output port

Transmission


A Proactive Routing Algorithm Classification

Non-adaptive algorithms(static)

Adaptive algorithms(dynamic)


NON-ADAPTIVE ROUTING


Non-adaptiveAlgorithms

Fixed Directory routing AKA static routing Manual configuration

(Selective) floodingand derivates


FixedDirectoryRouting

B C DA

RoutingTable of BA,E C,D

RoutingTable of A

B,D E,C ↓

RoutingTable of DA,B,C,E

RoutingTable of CA,B D E ↓

ERoutingTable of EA,B C,D


Pros and Cons

Administratorhas full control

Error prone It does not adapt

to topology changes


BackupRoute LoadBalancing

B C DA

RoutingTable of BA,E,C,DC,D,E,A

RoutingTable of BA,E,C,DC,D,E,A

RoutingTable of A

B,D,C,EE,C,B,D ↓

RoutingTable of A

B,D,C,EE,C,B,D ↓



RoutingTable of CA,B D

E,A,B ↓

RoutingTable of CA,B D

E,A,B ↓

ERoutingTable of EA,B C,D,B

RoutingTable of EA,B C,D,B


Static vs. Dynamic

CoreDynamicrouting

EdgeStatic routing

The only way to the

rest of network

Small routing tables

Multiple alternative routes


DYNAMICROUTING


Adaptive Algorithms

Centralized Routing Isolated Routing Distributed Routing Distance Vector Link State


Centralized Routing

Routing Control Center(RCC)

Calculates and distributesrouting tables

Needs informationfrom all nodes


Centralized Routing

Optimizes performance

Simplifies troubleshooting

Significant network loadin proximity of RCC


Centralized Routing

RCC is single pointof failure

RCC is bottleneck

Not suitable for highlydynamic networks


Isolated Routing

Each node decidesindependently

No exchange of information E.g., Backward Learning IEEE 8O2.1D bridges


Distributed Routing

Routers co-operatein exchanging connectivity information

Each router decides independently, but coherently

Combines advantages of isolated and centralized

routing


DISTANCEVECTOR ROUTING

ALGORITHM


Basic Principle

A

C

FE

2. A at dist 0and E at 1(from A1)

2. F at 0 andE at 1 (from

F1)

A1

F1

4. C at 0, A at 1(from C1), F at 1

(from C2), and E at 2 (from C1 and C2,

same dist)C1

C21. E at dist

0

3. A at 0 andE at 1

3. F at 0 andE at 1


Distance Vector

List of reachabledestinations (all!)

Distance from announcingrouter

Generated by each router Received from neighbors


Sample Scenario

A B

ED

Loc (A)A, 0

DV (B)A, 1B, 0C, 1D, 1E, 1

DV (D)A, 1B, 1C, 2D, 0E, 1

Routing informationstored by A

C


Distance VectorMergingand Generation

A B

ED

Loc (A) DV (B) DV (D)

A1

ROUT. TABLE (A)

A, 1B, 0

C, A1, 2C, 1D, 1

B, 1C, 2D, 0

A, 0B, A1, 1

CA2

E, 1 E, 1

A, 1 A, local, 0

D, A2, 1E, A2, 2

DV (A)

A, 0B, 1C, 2D, 1E, 2

Received from line A1Received from line A2


A B

ED

C

Loc (A) DV (B) DV (D) ROUT. TABLE (A)

A, 1B, 0C, 1D, 1

B, 1C, 2D, 0

A, 0

E, 1 E, 1

A, 1 A, local, 0

D, A2, 1E, A2, 2

DV (A)

A, 0B, 2C, 3D, 1E, 2

B, A2, 2C, A2, 3

Topology ChangeA1

A2


Example: Cold Start

A B

ED

A1

CA2

B1B2

B4

B3C1

C2E1

E2

E3

D1

D3D2

RT (A)A,loc,0

RT (B)B,loc,0

RT (C)C,loc,0

A sends its DV

B and D send their DVs

RT (D)D,loc,0

RT (E)E,loc,0

RT (A)A,loc,0

RT (B)A,B1, 1B,loc,0

RT (D)A,D1, 1D,loc,0

RT (C)C,loc,0

RT (E)E,loc,0

RT (A)A,loc,0B,A1, 1D,A2, 1

RT (B)A,B1, 1B,loc,0D,B2, 1

RT (D)A,D1, 1B,D2, 1D,loc,0

RT (C)A,C1 ,2B,C1 ,1C,loc,0

RT (E)A,E2, 2B,E2, 1D,E1, 1E,loc,0

All send their DVs. . .

RT (A)A,loc,0B,A1, 1C,A1, 2D,A2, 1E,A2, 2

RT (B)A,B1, 1B,loc,0C,B4, 1D,B2, 1E,B3, 1

RT (D)A,D1, 1B,D2, 1C,D2, 2D,loc,0E,D3, 1

RT (C)A,C1 ,2B,C1 ,1C,loc,0D,C2, 2E,C2, 1

RT (E)A,E2, 2B,E2, 1C,E3, 1D,E1, 1E,loc,0


Issues

Several problems Black Hole Count to infinity Bouncing Effect (loop)

Instability


Count to InfinityA BA1 CB1 B2 C1

Loc (B)B, 0

DV (A)A, 0B, 1C, 2

DV (C)A, 2B, 1C, 0

IS B:Loc (C)C, 0

DV (B)A, 1B, 0C, 1

RT (B)A,B2 ,3B,loc,0C,B2 ,1




Loc (C)C, 0

DV (B)A, 3B, 0C, 1

B sends DV

Loc (B)B, 0

DV (C)A, 4B, 1C, 0

C sends DV

Count to Infinity!

IS C:


Bouncing Effect

A BA1 CB1 B2 C1

Loc (B)B, 0

DV (A)A, 0B, 1C, 2

DV (C)A, 2B, 1C, 0

IS B:Loc (C)C, 0

DV (B)A, 1B, 0C, 1




Loc (C)C, 0

DV (B)A, 3B, 0C, 1

B sends its DV

IS C:


Issues

Partial solutions Split Horizon Path Hold Down Route Poisoning


Split Horizon“If C reaches destination A through B, it is useless for B trying to reach A

through C”

A BA1 CB1 B2 C1

Loc (B)B, 0

DV (A)A, 0

DV (C)C, 0D, 1

IS B:RT (B)B,loc,0C,B2 ,1D,B2, 2

DC2 D1

DV (CC1)C, 0D, 1

DV (CC2)A, 2B, 1C, 0


Split Horizon Prevents loops between twonodes

Speeds up convergence “Personalized” DVs toneighbors DV of C to B does not

contain destinationsreached through B

In actual implementations,route has to expire


Split Horizon on Mesh

A B

CD

A1 B1

B2C1

C2

D1B3

D2

Loc (B)B, 0

DV (A)A, 0

DV (C)C, 0D, 1


DV (D)C, 1D, 0

RT (D)A,D1, 2B,D1, 1C,D2, 1D,loc,0

Loc (D)D, 0

DV (B)A, 1B, 0C, 1

IS D:DV (C)A, 2B, 1C, 0

RT (C)A,C1, 2B,C1, 1C,loc,0D,C2, 1

Loc (C)C, 0

DV (B)A, 1B, 0D, 1

IS C:DV (D)A, 2B, 1D, 0

RT (C)A,C2, 3B,C1, 1C,loc,0D,C2, 1

RT (D)A,D2, 3B,D1, 1C,D2, 1D,loc,0

Loc (C)C, 0

DV (B)B, 0D, 1

IS C:DV (D)A, 2B, 1D, 0

Loc (D)D, 0

DV (B)B, 0C, 1

IS D:DV (C)A, 2B, 1C, 0

B sends its DV


B2C1

Split Horizon on Mesh

A B

CD

A1 B1

C2

D1B3

D2

Loc (B)B, 0

DV (C)A, 3C, 0D, 1

IS B:DV (D)A, 3C, 1D, 0

Loc (C)C, 0

DV (B)B, 0D, 1

IS C:DV (D)B, 1D, 0

Loc (D)D, 0

DV (B)B, 0C, 1

IS D:DV (C)B, 1C, 0

RT (C)A,C2, 3B,C1, 1C,loc,0D,C2, 1

Loc (C)C, 0

DV (B)B, 0D, 1

IS C:DV (D)A, 2B, 1D, 0

RT (D)A,D2, 3B,D1, 1C,D2, 1D,loc,0

Loc (D)D, 0

DV (B)B, 0C, 1

IS D:DV (C)A, 2B, 1C, 0

RT (B)A,B3, 4B,loc,0C,B2 ,1D,B3, 1

RT (C)B,C1, 1C,loc,0D,C2, 1

RT (D)B,D1, 1C,D2, 1D,loc,0

(from the previous slide)

C and D send their DVs


Path Hold DownIf link L fails, all destinations reachable through link L are

considered unreachable for a certain period of timeI.e., no routes to them are

computedA BA1 CB1 B2 C1

Loc (B)B, 0

DV (A)A, 0B, 1C, 2D, 3

DV (C)A, 2B, 1C, 0D, 1


DC2 D1

Quarantine!Loc (C)C, 0

DV (B)B, 0C, 1D, 2

DV (D)A, 3B, 2C, 1D, 0

IS C:RT (C)A,C2, 4B,C1, 1C,loc,0D,C2, 1

Count to Infinity!


Path Hold Down High convergence time forthe examined node (evenwith an alternative path)

The router that noted thefault may not participateto any loop at least untilthe timeout of Hold Downtimer


React on Cost IncreaseRouting loops happen when

routes have increasing costs

Cost-increasing routes inDVs are not used Two subsequent

advertisements show a costincrease

Possibly with Path Hold Down Might block routes with

legitimate cost increase


Route PoisoningAn invalid route is

advertised at infinite distance

Instead of just omitting it It would have to expire Faster convergence time E.g., when link fails orcost increases

It can substitute orcomplement Path Hold Down


Split Horizon with Poisonous Reverse More aggressive No theoretical advantages Practically, no need towait for route expiration

A BA1 CB1 B2 C1

Loc (B)B, 0

DV (A)A, 0B, infC, infD, inf

DV (C)A, infB, infC, 0D, 1


DC2 D1

DV (CC1)A, infB, infC, 0D, 1

DV (CC2)A, 2B, 1C, 0

D, inf


The Bottom LineRouters do not knowthe network topology

Based on distance vectorsB cannot distinguish

A B C

A B C


Advantages

Simple to implement Protocols simple to deploy Very little

configuration


Shortcomings

Exponential worst casecomplexity and convergencetime O(n2) to O(n3)


Shortcomings

Convergence limitedby slower linksand routers set pace

Complex tuning


Shortcomings

Complex troubleshooting Large routing traffic

(and storage)

Not suitable for large complex networks


Path VectorEliminates routing loops

IS A:

A BA1 CB1 B2 C1 DC2 D1

Loc (A)A, 0

PV (B)A, 1, [B]B, 0, [-]C, 1, [B]D, 2, [B,C]

PV (C)A, 1 [C]B, 1, [B]C, 0, [-]D, 1, [D]

RT (A)A, loc, 0B, A1, 1C, A2, 1D, A2, 2

PV (A)A, 0, [-]B, 1, [A]C, 1, [A]D, 2, [A,C]

A2


LINK STATE ROUTING ALGORITHM


Basic PrinciplesA

C

FE

1. I, nodeE, connectto A and F

2. I, node A,connect to E

and C

3. I, node F,connect to E and

C

E F

A

E

C

F

A

E

C

F

A


Basic Principles

Information on the stateof each link Link state

A local map Sent by each node

to all other nodes Selective flooding


Basic Principles

Nodes build a network map The same on all nodes

Each node computes routeson the map Dijkstra (shortest pathfirst) algorithm


Link State DatabaseA B

ED

LS (A)B, 1D, 1

LS (B)A, 1C, 1D, 1E, 1

LS (D)A, 1B, 1E, 1

Stored by each router

C

LS (C)B, 1E, 1

LS (E)B, 1C, 1D, 1


Dijkstra Algorithm

Low complexity

L log (N)L: number of linksN: number of nodes

Shortest Path First


Shortest Path First The next node “nearest” tothe root is identified

Its path is inserted intothe routing table


ExampleBA

C D

E

2

1 3

3

2

3

root PATHTEMP

F

2

2)BA

C D

E F

3)BA

C D

E F

4)BA

C D

E F

5)BA

C D

E F

6)BA

C D

E F

BA

C D

E F

7)BA

C D

E F

8)BA

C D

E F

9)BA

C D

E F

10)

1)


Rapid Convergence

LSs spread quickly No intermediateprocessing


Limited Routing Traffic and Storage

Link states are small Fast and efficientneighbor greeting


Other Advantages

It rarely generates loops Simple to understandand troubleshoot All nodes haveidentical databases


Shortcoming

High implementationcomplexity Selective flooding First implementationtook several years

Protocols with complexconfiguration


Link State GenerationIn principle: when thereis a topology change

Actual protocols: LS aregenerated periodicallyIncreased reliability

routing and forwarding - polito

Documents