Introduction to MANET Routing - nada.kth.se · Introduction to MANET Routing Laura Marie Feeney...

Post on 07-Apr-2018

226 views 2 download

Transcript of Introduction to MANET Routing - nada.kth.se · Introduction to MANET Routing Laura Marie Feeney...

lmfeeney@sics.se

Introduction to MANET Routing

Laura Marie FeeneySwedish Institute of Computer Science

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

MANET

MANET: Mobile Ad hoc NETworkn mobile wireless network, capable of

autonomous operationn operates without base station

infrastructuren nodes cooperate to provide

connectivityn operates without centralized

administrationn nodes cooperate to provide services

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Ad hoc routing

infrastructureless network

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Ad hoc routing

“equivalent” topology

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Ad hoc routing

dynamic multihop routing

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Ad hoc routing

route failure due to mobility

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Ad hoc routing

dynamic route repair

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Applications

n tactical (military) networks (FOI)n disaster recovery servicesn metropolitan/campus-area

communication networks (UU)n sensor networks (SICS)n enhanced cellular networks (KTH)n delay-tolerant networking (LUTH)

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Definition

How is a MANET different from othernetworks?n Internetn WLANn MobileIP

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Internet/Intranet

default router

Internet routing(+ humans)

networkprovider

managing infrastructure requiressignificant expert configuration

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Cellular (WLAN)

mobiles communicate only withbase-stations

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

MobileIP

correspondent

CN

MNhome

FA

Mobile-IP allows a node to change itspoint of attachment to the network

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

MobileIPCN FA

HA

MN

mobiles register location with homeagent, which tunnels traffic

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Status

Still largely an R&D activityn IETF MANET working group

u two protocols will become“proposed standards”

n IRTF working groupn research communityn small-scale testbeds and simulation

experiments

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Overview

n ad hoc routing problemu challengesu design choicesu protocol example

n other problemsu security & cooperationu servicesu wireless issues

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Challenges

n distributed state in unreliableenvironment

n dynamic topologyn limited network capacityn wireless communication

u variable link qualityu interference and collisionsu energy-constrained nodes

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Criteria

n effectivenessu convergence/recoveryu scalability (number of nodes,

density)n performance

u data throughputu route latency (delay)u route optimality

(hops/stability/diversity)u overhead cost

(packets/bandwidth/energy)

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Alphabet Soup

many proposed protocols:AODV CEDAR ABR FSR

TORA GSR OLSR LANMAR

ZRP LAR DSR OSPF++

RDMAR CBRP DSDV WRP

TBRPF CGSR GPSR

protocols in red are best known

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Design Choicesprotocols fall into a few main categoriesn

n on-demand (reactive)n table-driven (pro-active)n flooding-basedn cluster-basedn geographicn application specific (cross-layer)

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Design Choices

no pre-assigned backbonecan designate a backbone dynamicallybackbone provides structure for thenetworkn increases (?) scalabilityn cost to maintain backbone structuren disproportionate load on backbone

nodesCEDAR is an example

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Routing backbone

12

13

1

4

58 11

10

97

6

3

142

n connected backbone; each node hasa backbone neighbor

n distributed computation of aconnected minimum dominating set ishard

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Routing backbone

?

2

13

1

4

58 11

10

97

6

3

1412

?

n maintaining the backbone can becostly

n common strategyperiodic broadcast of neighbor data,backbone nodes self-nominate viaadaptive backoff

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Routing backbone

12

13

1

4

58 11

10

97

6

3

142

n non-backbone nodes have “default”router

n how to route across the backbone??

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Reactive Routing

n reactive (on-demand) protocoln only obtain route information when

neededn advantages

u no overhead from periodic updaten disadvantages

u high route latencyu route caching can reduce latency

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Pro-active routing

n pro-active (table-driven) protocoln more similar to conventional routingn advantages

u low route latencyu state information

n disadvantagesu high overhead (periodic table

updates)u route repair depends on update

frequency

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (DYMO)

Ad hoc On-demand Distance Vector(Perkins et.al.)

n conventional distance vectoru nodes exchange distance tables

with their neighborsu periodic exchange and immediate

update for changesu routing table selects shortest path

n exchange a lot of information that isnever used

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV Strategy

n on-demand variant of conventionaldistance vector

route request (RREQ) is floodedthrough the networkroute discovery creates (temporary)reverse routesroute reply (RREP) activatesforward route

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV Strategy (cont’)

n handling topology changelink failure generates route error(RERR)destination managed sequencenumber ensures loop freedom

n simplified presentation follows...

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV

14

13

12

1

4

58 11

10

97

6

32

n on-demand routing protocoln node 1→ 14

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RREQ)

14?

214

1

13

129

10

11

7

8

6

5

4

3

n broadcast flood route requestmessageu (broadcast traffic in red)

n “wireless multicast advantage”

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RREQ)1/14: 1

14

1

13

129

10

11

7

8

6

5

4

3

1/14: 1

1/14

: 12

n node from which RREQ was receiveddefines a reverse route to the sourceu (“reverse routing table entries” blue)

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RREQ)

1/14: 1

14?

14?

3 14?

4

5

6

8

7

11

10

912

13

1

142

n route request is flooded through thenetwork

n reverse routing table entries (blue arrows)

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RREQ)

1/14: 1

3

1/14:5

4

5

6

8

7

11

10

912

13

1

142

n unreliable communicationn destination managed sequence

number, ID prevent looping

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RREQ)

10

4

5

6

8

7

11

3

912

13

1

142

n flooding is expensiven broadcast collision problem

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RREQ)

10

4

5

6

8

7

11

3

912

13

1

142

n route request arrives at thedestination

n two routes are discovered

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RREP)

10

4

5

6

8

7

11

3

912

13

1

142

n destination sends route reply (setsequence number)u (unicast reply in magenta)

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RREP)

14:14(n)

10

4

5

6

8

7

11

3

912

13

1

142

n routing table now contains forwardroute to the destinationu (“reverse routing table entries” in blue)

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RREP)

10

4

5

6

8

7

11

3

912

13

1

142

n route reply follows reverse route backto the source

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RREP)

14:13(n)

10

4

5

6

8

7

11

3

912

13

1

142

n setting the forward routing tableentries along the way

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RREP)

10

4

5

6

8

7

11

3

912

13

1

142

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RREP)

10

4

5

6

8

7

11

3

912

13

1

142

n route reply reaches the source

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RREP)

10

4

5

6

8

7

11

3

912

13

1

142

n source adopts destination sequencenumber

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV

3

14

1

13

129

10

11

7

8

6

5

4

2

n traffic flows along the forward routen forward route is refreshed, reverse

routes time out

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RERR)

32

14

1

13

129

10

11

7

8

6

5

4

n link failure detection

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RERR)

14:10(n+1)

214

1

13

129

10

11

7

8

6

5

4

3

n return error message to the source(increment sequence number)

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RERR)

214

1

13

129

10

11

7

8

6

5

4

3

n source receives route error

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

AODV (RREQ)

3

14

1

13

129

10

11

7

8

6

5

4

2

n re-initiates route discovery process

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Criteria

n effectivenessconvergence/recoveryscalability (number of nodes,density)

n performancedata throughputroute latency (delay)route optimality(hops/stability/diversity)overhead cost(packets/bandwidth/energy)

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Design Choices

n reactive (on-demand) protocolu high route latencyu no overhead from periodic updateu route caching can reduce latency

n pro-active (table-driven) protocolu low route latencyu high overhead (periodic table

updates)u route repair depends on update

frequency

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

OLSR

Optimized Link State RoutingJacquet et. al.

n conventional link-state routingu beacon to determine neighborsu for each node, disseminate its links

to all other nodesu use SPF algorithm to generate

routing tablen high overhead, exchange information

for links that are never used

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

OLSR Strategy

n optimized variant of conventional linkstate routing

for each node, disseminate onlysome of its linksfor each node, only disseminateinformation received via some linksuse SPF algorithm to generaterouting table

n “some (carefully selected!) links” =multipoint relay set

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

2-hop Neighborhoodbroadcast periodic “hello” messages

each message contains a list ofneighborseach node discovers its 2-hopneighborhooddiscovers failed linksdiscovers bi-directional links

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Bi-directional Links

HELLO(3)={1,2,4,7}

HELLO(6)={2,3,7}

6

4

3

14

1

2

13

129

10

11

7

85

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Bi-directional Links

NBR(3)={1,2,4,7}

NBR(3)={1,2,4,7}HELLO(6)={2,3,7}

5

3

14

1

2

13

129

10

11

7

8

6

4

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Multipoint Relay

multipoint relay set (MPR): subset of anode’s 1-hop neighbors, such that eachof its 2-hop neighbors is a 1-hopneighbor of a node in the MPR setin practice, approximate optimal MPRsetnote that each node independentlydetermines its own MPR set (no global“network MPR set”)

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Multipoint Relay

122

6

9

10

1185

3

7

4

14

13

1

one and two hop neighbors of node 4

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Multipoint Relay (MPR set)

MS(3)={4,...}

1

2

6

9

10

1185

3

7

4

14

13

12

MPR(4)={3,7,8}

MS(8)={4,...}

MS(7)={4,...}

node 5 is not needed in the multipointrelay set

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Dense Network

12

1

2

6

9

10

1185

3

7

4

14

13

with greater node density, theproportion of relay nodes is smaller

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Dense Network (MPR set)

{3,8}

{4,10}

{10}

{1,4,8}

{8,10}

{4,7}{1,3}

{3,5}

{7,8,12} {9,10}

{10}{7,8}

{6,4,10}

{2,7}

9

13

14

1

511

10

127

32

4

8

6

nodes which are not in the MPR setare somehow redundant

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

Dense Network (MS set)

9

13

14

1

511

10

127

32

4

8

6

multipoint selector (MS) set is theinverse of MPR set

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

OLSROperation:

each node uses HELLO message tocalculate and announce its MPR seta node sends link state informationonly for nodes in its MS set (for whichit is an MPR)each node processes (SPF routes) alllink state messagesa node only rebroadcasts link statemessages from nodes in its MS set

SICS

SWEDISHINSTITUTE OF

COMPUTERSCIENCE

lmfeeney@sics.se

OLSR (Dense Network)

only disseminate link data for greennodesonly rebroadcast data from green

nodes

1: 4 2 3 5 2: 1 3 6 3: 1 2 4 6 7

4: 1 3 5 7 8 5: 1 5 8 6: 2 3 7

7: 3 4 6 9 10 8: 4 5 9 10 11 9: 7 10 12

10: 7 8 9 11 12 13 11: 8 10 13 12: 9 10 13 14

13: 10 11 12 14 14: 10 12 13