PRIN 2005 - WOMEN PROJECTResearch Unit: University of Naples Federico II

G. Ferraiuolo

giancarlo.ferraiuolo@unina.it

T3.6: Location management and routing in wireless ad hoc and mesh networks

Rome, 19.01.07 WOMEN – 3° Meeting 3

Mobility management

• WMNs mesh routers have low mobility:– monitoring the connectivity

• Mesh clients are mobile:– Mobility management is essential for the ad hoc

domain

• Mobility management: location and handoff management

Rome, 19.01.07 WOMEN – 3° Meeting 4

Location management in MANETs

• Node’s identifier and node’s location can not be statically related in self organizing networks

• Location is retrieved and then routing is performed

• Routing based on location information can improve scalability respect to proactive/reactive approaches

Rome, 19.01.07 WOMEN – 3° Meeting 5

DHT based location management

• Distributed• Dynamic management• Low control message overhead• P2P applications for locate information (scalability and

resilience to failures)• DHT based routing:

– Location by GPS or other dedicated systems

– Topology location (by dynamic addressing)

Rome, 19.01.07 WOMEN – 3° Meeting 6

Addressing in MANETs

• Twofold goal:– Assign a like-IP address in mobile environments– Solve the location management problem for

supporting routing

• Mandatory requirements:– Distributed mechanism– Scalability

Rome, 19.01.07 WOMEN – 3° Meeting 7

Resume

• There are 3 problems to optimize:

– Dynamic Address Allocation– Dynamic address based Routing– Distributed Hash Table Lookup

• There is still work to do toward the definition of a realistic scalable routing approach based on DHT and Dynamic Addressing

Rome, 19.01.07 WOMEN – 3° Meeting 8

Distributed Address Allocation

• Identifier: static (IP)

• Routing address: dynamic

• Each node can assign a valid address to a joining neighbor node

• The assigning node also delegates control over half of its address space to the requester

• Problems: network merge and partition yield to address duplication

Rome, 19.01.07 WOMEN – 3° Meeting 9

Distributed Address Allocation

A level-k sibling of a given address is defined as the subtree that share the same immediate parent of the level-k subtree of the considered address.

Example: sibling(100) = 101, 11X, 0XX

Rome, 19.01.07 WOMEN – 3° Meeting 10

Distributed Address Allocation

• Each node build a state table– The table stores an entry for each one of node’s level-k

sibling subtree– If an entry is empty, the new node get an unoccupied

address in that subtree– The new node chooses the largest unoccupied

address set– Table dimension O(log n)

• The table is used also for routing

Rome, 19.01.07 WOMEN – 3° Meeting 11

Distributed Address Allocation

• Related works:– Stateless approach: flood the network for duplicate

address detection– Stateful approach (DART): based on the underlying

Neighbor Discovery Protocol

• Other stateful approaches dose not face effectively network partition problem– We use the identifier to detect partitions and solve the

contention

Rome, 19.01.07 WOMEN – 3° Meeting 12

Routing

• Location information is embedded in the dynamic address:

– Hierarchical distance-vector routing based on dynamic address

– A node compare the address with its own, if the i-th bit is different, it forwards the packet toward the i-th sibling

– Routing is a recursive descending through the address tree

– Prefix subgraph constraint assures robustness

Rome, 19.01.07 WOMEN – 3° Meeting 13

Prefix subgraph constraint

• Nodes with a given address prefix form a connected graph

• The longer the shared address prefix between two nodes, the shorter the distance in the topology

• Routing entries for distance nodes can remain valid despite local topology changes

• Reduce size of routing tables and updates

Rome, 19.01.07 WOMEN – 3° Meeting 14

Routing optimization 1

• Tree address structure– simple and manageable (+)– low route selection flexibility for routing (-)

• Optimization opportunities:– Enhancing address capability (-)– Increasing table dimension (+)

• Additional neighbor information in the routing table– Low increment of the table size, same routing overhead

Rome, 19.01.07 WOMEN – 3° Meeting 15

Routing optimization 1

Rome, 19.01.07 WOMEN – 3° Meeting 16

Routing optimization 2

• Toward an effective routing protocol:– Cross layering is mandatory in MANETs– A simple interaction with MAC 802.11 can be very useful– When a link failure is detected, the corresponding entry is

removed from the state table– Routing is attempted using another available entry

• Resuming: – Routing protocol 1 (R1): minimum table size – Routing protocol 2 (R2): neighbor information – Routing protocol 3 (R3): minimum table size with cross layer– Routing protocol 4 (R4): neighbor information with cross layer

Rome, 19.01.07 WOMEN – 3° Meeting 17

Simulated scenario

• Network Simulator 2

• Mobility model:– random waypoint;– speed [0.5;5] m/s, pause time [0;100] s;– density ≈ 121 nodes/km2;

• Data traffic model:– Constant Bit Rate (CBR) on UDP;– throughput CBR = 0.25 throughput link

Rome, 19.01.07 WOMEN – 3° Meeting 18

Simulation results

0 50 100 150 200 250 300 350 400 450 5001

2

3

4

5

6

7

# Nodes

# H

op

sHops

R1R2

Rome, 19.01.07 WOMEN – 3° Meeting 19

Simulation results

0 100 200 300 400 5000

100

200

300

400

500

600

700

800

900Delay

# Nodes

De

lay

(mse

c)

R1R2

Rome, 19.01.07 WOMEN – 3° Meeting 20

Simulation results

0 50 100 150 200 250 300 350 400 450 5000

10

20

30

40

50

60

70

80

90

# Nodes

# T

ab

le e

ntr

ies

Table Dimension

R1R2

Rome, 19.01.07 WOMEN – 3° Meeting 21

Simulation results

0 100 200 300 400 50020

30

40

50

60

70

80

90

100

# Nodes

% P

ack

et d

eliv

ery

Ra

tioPacket Delivery Ratio

R1R3R2R4

Rome, 19.01.07 WOMEN – 3° Meeting 22

Conclusions

• Preliminary results:– Address allocation scheme is effective in

assigning valid addresses – Better performance in routing is achieved with

slightly more complexity– Cross layering is necessary to achieve

effective performance