Scalable Team Multicast in Wireless Ad hoc networks Exploiting Coordinated Motion

Mario GerlaUniversity of California, Los

Angeles

Introduction Many team-orientedteam-oriented operations in MANET

scenarios Search and rescue, disaster relief operation,

battlefields Each team tends to move together (affinity team

model) a chosen node (e.g., landmark) can represent a team

Often, all nodes or none in a team join a multicast group

Affinity team model simplifies mobility handling, and thus allows a scalable multicast protocol design

LANMAR (Landmark ad hoc routing protocol) works well with affinity team model

Introduction (2) Proposed idea, two-tier team multicast, team multicast, called

M-LANMARM-LANMAR Unicast tunneling from the sources to the

representative of each subscribed team Flooding within a team

Advantages of M-LANMAR High reliability via unicast tunneling and

flooding Easy to manage the large networks based

on motion affinity model

LANMAR (Landmark Ad Hoc Routing Protocol) –

Background

Proactive Routing Efficient handling affinity team model Using the notion of landmarks to keep track of logical

subnets (teams) Using two routing schemes

A local proactive routing: within a limited scope, nodes exchange their routing table each other

A long haul distance vector routing: a landmark of each subnet is propagated to the whole network

Routing Tables Local routing table Landmark table

LANMAR-Example

Logical SubnetLogical Subnet

LandmarkLandmark

node1node2

LM1 LM2

LM3

Landmark Table of node1(subnet_addr, lm_addr, nextHop, …)

Node3 (src)

node3LM3subnetAddr3

node3LM2subnetAddr2

node2LM1subnetAddr1

Local Routing Table of node1(destAddr, nextHop, …)

…

Node3Node3

Node2Node2

dest

Long haul routinglocal routing

M-LANMAR(Multicast-enabled LANMAR)

Extension of LANMAR Proactive scheme Supporting both unicast and multicast routing

with very low extra overhead Scalable as the network size and number of

groups increase Join MC group: advertising by piggybacking

subscribed multicast groups IDs to landmark broadcast packets

Leave MC group: simply not advertising/timeout The sources can find joined teams in their

landmark table

Source node

LM1 LM2

LM3

Subscribed Teams

LM4

Tunneling

Flooding

Flooding

Scope = 2

Scope = 2

(MC1)

MC1

MC1

node3LM2subnetAddr3

node3LM4subnetAddr2

node2LM1subnetAddr1

Landmark table of source node

…

Simulation Compared with

ODMRP (On Demand Multicast Routing Protocol) Flooding

Environments QualNet 2.9 Each source generates data in a CBR fashion Transmission range: 376m, bandwidth 2Mbits Network size: 6000 x 6000 m2

Nodes: 1000 nodes into 36 teams Mobility: 2 m/s with 10 seconds pause time

Following “Reference Point Group Mobility” Packet size: 512 bytes

Simulation Results – Static Network (1000 nodes, 3teams for each

group, 1pkts/sec)

As the number of multicast groups increases ODMRP suffers from large control overhead and

collisions M-LANMAR achieves high delivery ratio (by unicast

tunneling and flooding)

Number of MC Groups(#) Number of MC Groups(#)

Delivery

Ratio

Norm

alized CT

RL

O

H

Simulation Results – Mobile Network(1000 nodes, 3teams for each group, 2m/s, 10p)

4pkts/secODMRPM-LANMARFLOOD

1pkt/secODMRPM-LANMARFLOOD

Simulation Results – Mobile Network(1000 nodes, 3teams for each group, 2m/s, 10p) (2)

As the number of multicast groups increases With small number of groups, ODMRP

outperforms M-LANMAR because of its mesh-based multicast structure (redundant paths); M-LANMAR potentially experiences many link breakages

With large number of groups, flooding suffers due to heavy overhead

With high offered load (right), M-LANMAR outperforms ODMRP (ODMRP suffers from heavy redundant forwarding)

Reliable Multicast Support Reliable Adaptive Lightweight Multicast (RALM)

Targeting general multicast protocols NACK-Oriented mechanism Rate-based congestion control

Send-and-wait mechanism (freeze the sender’s buffer upon receiving a NACK); congestion handling

Round-robin recovery Sources recover the lost pkts for each NACKER one at a

time in a round-robin fashion Prevent ACK implosion

Combining with M-LANMAR Only landmarks (say representatives) send feedback

(e.g. NACK/ACK) to the source Prevents unnecessary feedback implosion

Simulation Results with RALM (1000 nodes, 3teams for each group, 2m/s, 10p, 5 multicast groups) –Delivery ratio

Delivery Ratio

0

0.2

0.4

0.6

0.8

1

1.2

512 1024 1280 1689.6 2560 5120

Offered Load (Bytes/sec)

Deliv

ery

Ratio

M- LANMAR w/ UDP ODMRP w/ UDPM- LANMAR w/ RALM ODMRP w/ RALM

•Same parameters to the mobile network experiment•Increase the offered load (number of pkts/sec)•ODMRP suffers from feedback implosion

Simulation Results with RALM (1000 nodes, 3teams for each group, 2m/s, 10p, 5 multicast groups) –Throughput

Throughput

0

5000

10000

15000

20000

512 1024 1280 2560 5120

Offered Load (Bytes/sec)

Thr

oug

hput

(Bits

/sec

)

M- LANMAR w/ UDPODMRP w/ UDPM- LANMAR w/ RALMODMRP w/ RALM

Discussions Possible extensions of M-LANMAR

Efficient resource discovery via content based multicast

Use existing MANET multicast protocols to multicast a packet to all landmarks

Scalable as the number of teams increases Provide a congestion controlled reliable

transport layer like TCP between the sources and the landmarks

Conclusion Propose a new multicast paradigm

Team multicast Design M-LANMAR as an example of team

multicast Study the performance of M-LANMAR compared

with ODMRP and FLOOD Flat multicast has scalability limitations M-LANMAR provides an efficient platform for a reliable

and congestion controlled multicast protocol (e.g., TCP) Apply a reliable transport protocol, RALM over M-

LANMAR and ODMRP M-LANMAR is an efficient platform for a reliable multicast

protocol

Thank you

Any Questions?