Lecture 3-1: Networking Architecture, Routing Protocols and Algorithms

Lecture 3-1: Lecture 3-1: Networking Networking

Architecture, Routing Architecture, Routing Protocols and Protocols and

AlgorithmsAlgorithms

Lecture 3-1: Lecture 3-1: Networking Networking

Architecture, Routing Architecture, Routing Protocols and Protocols and

AlgorithmsAlgorithms

A Cluster-based architecture A Cluster-based architecture for Dynamic Sensor Radio for Dynamic Sensor Radio

NetworksNetworks

A Cluster-based architecture A Cluster-based architecture for Dynamic Sensor Radio for Dynamic Sensor Radio

NetworksNetworks11Jiro UchidaJiro Uchida

2 2 Islam A.K.M. MuzahidulIslam A.K.M. Muzahidul33Yoshiaki KatayamaYoshiaki Katayama

44Wei Chen Wei Chen 55Koichi WadaKoichi Wada

1,2,3,5 1,2,3,5 Nagoya Institute of TechnologyNagoya Institute of Technologyandand

44 Tennessee State University, USATennessee State University, USA

International Conference on System Sciences, 2006

OutlineOutline• Sensor Radio NetworksSensor Radio Networks• Cluster-based sensor radio network Cluster-based sensor radio network

architecture architecture • Broadcasting Broadcasting • Construction of Cluster-based sensor Construction of Cluster-based sensor

radio networkradio network dynamicallydynamically– node-move-in algorithm– node-move-out algorithm

• Simulation resultsSimulation results• Conclusion Conclusion

A sensor radio network is a collection of sensor nodes, where each sensor node has a sensor array, a controlling processor and transmitter-receiver communication device.

What is Sensor Radio What is Sensor Radio Networks ?Networks ?

vxu

u,v,xu,v,x:: communicationcommunication devicesdevices

Transmission range of devices

A sensor radio network is a collection of sensor nodes, where each sensor node has a sensor array, a controlling processor and transmitter-receiver communication device.

uv

x

Transmission range of nodes

u,v,xu,v,x:: communicationcommunication devicesdevices

What is Sensor Radio What is Sensor Radio Networks ?Networks ?

• Each node works per synchronized Each node works per synchronized round by a global clock (like GPS).round by a global clock (like GPS).– Round: Discrete time step of the global

clock.

• Actions of nodes per Round:– transmission or reception– process before and after communication

Collision detection is not availableCollision detection is not available

distinguishable

Collision detection

collision

Each node cannot receive two or more messages in one roundEach node cannot receive two or more messages in one round

receivable unreceivable(collision)

• Each node represents a device.• Each directed edge uv represents that, a node u can transmit to a node v.

u

v

w

u

vw

v

u

u,v,w: devicesw

• We consider a sensor radio network bi-directional, i.e., the transmitting range of each device is the same .

Reachability graphReachability graph

• Reconfigurable networking architecture: A dynamic sensor radio network that supports two atomic operations: node-move-in and node-move-out

• Networking functions: broadcasting/multicast/unicast

７

６

5

43

1

2

９

８

Sensor Network ModelingSensor Network Modeling

Assumption:

•before a join: n1

•after a leave: the graph is connected

７

６

5

43

1

2

９

1０

８

1

6

10

Cluster

member

cluster edge

Cluster heads: Maximum independent Set (MIS)

Cluster-based architecture of a Cluster-based architecture of a sensor networksensor network

head

Backbone graphBackbone graph

７

６

5

43

1

1０

８

1

6

10

Cluster

member

cluster edge

Cluster

Cluster

Gateway edge candidates

Gateway edge

2

9

gateway node

2

9

head

Backbone TreeBackbone Tree BT(G)BT(G)Gateway nodeGateway node

cluster edge

Gateway edge

• The total number of nodes in BT(G) tree is at most 2p-1, p is the number of clusters.

• The number of the clusters p is less than pG, the smallest number of disjoint complete subgraphs in G.

Cluster-based network CNet(G)Cluster-based network CNet(G)

headhead

gatewaygateway

membermember

７

5

43

８

1

6

10

2

9

rootroot

What is broadcasting?

• Broadcast is dissemination of a message from one node to all nodes in the network.

７

６

5

43

1

2

９

８10

MMMM

MMMMMMM

Source nodeSource node

Some previous Some previous ResultsResults

Problem Information available to the

nodes

Completion rounds

Broadcasting

self ID only

whole network information

)(n

n : number of nodes in the network

D : diameter of G

)(nO

)(log2 n)log( 5 nDO

Construction of Cluster-based Cluster-based dynamic sensor radio networksdynamic sensor radio networks

Case 1 (1-hop info of CNET(G) and

G)

Case 2 (1-hop info of CNET(G))

•Information listed in case 2 and • set of all 1-hop neighbor nodes in G and their status (i.e., head or gateway or member)

• self ID• set of neighbor gateway nodes (if head)• set of neighbor head nodes (if gateway)• self parent (except the root)• self child (except the member nodes)

Our ResultsOur Results Information

available to the nodes

Problem Completion rounds

Case 1 and Case 2 Broadcasting Case 1 node-move-in expected

node-move-out

Case 2 node-move-in expected

node-move-out

expected

expec.

)(logqO

)(qO

|)(|TO

)log.|(| rTO

)( pO

subtree-T andnew of nodesneighbor ofnumber

. G in subgraph completedisjoint theofnumber

smallest the, thanless clusters theofnumber

q

pp G

)|(| tpTO

Broadcasting algorithm

M

M

A broadcasting from any node is done in O(p)- rounds.

In a dense graph our algorithm is more efficient, since ,

where n is the number of nodes in the network.

np

headhead

gatewaygateway

membermember

cluster edge

Gateway edge ７

5

43

８

1

6

10

2

9

rootroot

M

M

M

M

M

MM

M

M

Source node

Construction of Cluster-based Construction of Cluster-based sensor radio networksensor radio network

dynamicallydynamically

• node-move-in algorithm• node-move-out algorithm

When the nodes are organized with When the nodes are organized with partial 1-hop datapartial 1-hop data

• Procedure SelectWinner will be used in our node-move-in and node-move-out algorithms (randomized) of case 2 .– The average rounds requirement is O(log q).

７

5

43

８

1

6

10

cluster edge

Gateway edge

2

9

2

3

New nodeAddMeAddMeAddMe

node-move-in algorithmnode-move-in algorithm

• If the winner is a head node

I’mMember

11

headhead

gatewaygateway

membermember

７

5

43

８

1

6

10

cluster edge

Gateway edge 2

9

2

3

New node

• If the winner is a gateway node

• Send ChkHead message

• if there are more than one neighboring heads send SelectHead message

• Select one head and then send I’mMember message

11

headhead

gatewaygateway

membermember

I’mHead

７

5

43

８

1

6

10

cluster edge

Gateway edge

2

911

New node

• If the winner is a member nodeheadhead

gatewaygateway

membermember

11

3

BeGateway

• Joining in G can be done in expected rounds, where q is the number of neighbor nodes of new.

)(logqO

７

5

4

８

1

6

10

cluster edge

Gateway edge 2

9

2

3

• Send Leaving message

• All leaving message receiving node delete the node from their list.

If the leaving node is a member

leavingleavingleavingleavingleaving

Node-move-out algorithmNode-move-out algorithm

headhead

gatewaygateway

membermember

rootroot

７

5

43

８

1

6

10

cluster edge

Gateway edge 2

9

6

2

3

• The leaving node sends Leaving message

• All leaving message receiving node delete the node from their list and act accordingly.

If the leaving node is a head without child

leavingleaving

headhead

gatewaygateway

membermember

leaving node

9

Subtree H

If the leaving node is a head If the leaving node is a head (with child) or gateway(with child) or gateway

5

43

８

1

6

10

2

9

deldelheadhead

gatewaygateway

11

7

12

T- subtreeT- subtree

3

8

membermember

rootroot

del –leaving node del –leaving node

Subtree H

If the leaving node is a head If the leaving node is a head (with child) or gateway(with child) or gateway

5

43

８

1

6

10

2

9

deldel

headhead

gatewaygateway

11

7

12

T- subtreeT- subtree

3

7

11

9

8

10

membermember

rootroot

JoinMeJoinMeJoinMe

Token-B

If received JoinMe message in previous round

　　　 Step-1: if in T or has sent Token-B in previous round

　 send self id; 　

　　　 if JoinMe message sending node receives message (i.e. the node 　　　　　

　　　 does not have any neighbor in H)

　　　　　　 Continue Eulerian(T);

else

Goto step-2;

Send id

Subtree H

If the leaving node is a head If the leaving node is a head (with child) or gateway(with child) or gateway

5

43

８

1

6

10

2

9

deldelheadhead

gatewaygateway

11

7

12

T- subtreeT- subtree

3

8

membermember

rootroot

Token-B

JoinMeJoinMeJoinMeJoinMeStep-2: send message PerformSelectWinner;

if nodes that are not in T but received message PerformSelectWinner

call procedure SelectWinner;

After receiving winner’s id follow node-move-in-2 in order to finish join;

Send id

Send id

Subtree H

If the leaving node is a head If the leaving node is a head (with child) or gateway(with child) or gateway

5

43

８

1

6

10

2

9

deldelheadhead

gatewaygateway

11

7

12

T- subtreeT- subtree

BeGateway

3

7

11

6

9

6

8

10

membermember

rootroot

• Leaving of a node from G can be done in expected O( ｜ T| + r. log ∆) rounds, where r is the number of border nodes in T and ∆ is the number of nodes in subtree H that have edges with nodes in T.

Subtree H

A deterministic move-out algorithmA deterministic move-out algorithm

5

43

８

1

6

10

2

9

deldelheadhead

gatewaygateway

11

7

12

T- subtreeT- subtree

3

8

membermember

rootroot

del –leaving node del –leaving node

Subtree H

A deterministic move-out algorithmA deterministic move-out algorithm

5

43

８

1

6

10

2

9

deldelheadhead

gatewaygateway

11

7

12

T- subtreeT- subtree

3

8

membermember

rootroot

del –leaving node del –leaving node

Subtree H

If the leaving node is a head (with If the leaving node is a head (with child) or gatewaychild) or gateway

5

43

８

1

6

10

2

9

deldel

headhead

gatewaygateway

11

7

12

T- subtreeT- subtree

3

7

11

9

8

10

membermember

rootroot

Token-B

Subtree H

If the leaving node is a head If the leaving node is a head (with child) or gateway(with child) or gateway

5

43

８

1

6

10

2

9

deldelheadhead

gatewaygateway

11

7

12

T- subtreeT- subtree

3

8

membermember

rootroot

Token-B

Subtree H

If the leaving node is a If the leaving node is a head (with child) or head (with child) or

gatewaygateway

5

43

８

1

6

10

2

9

deldelheadhead

gatewaygateway

11

7

12

T- subtreeT- subtree

BeGateway

3

7

11

6

9

6

8

10

membermember

rootroot

• Leaving of a node from G can be done in O( ｜T| + p+ t) rounds, T is a subtree, p is the number of cluster, and t is the number of nodes in subtree H that are in 3-hop distance from the border nodes in T.

When nodes have more network When nodes have more network information i.e., total 1-hop datainformation i.e., total 1-hop data

• Joining in G can be done in expected time O(q) rounds, where q is the number of neighbor nodes of new.

• Leaving of a node from G can be done in O(|T|) rounds.

Simulation Results

• Simulation is performed on random unit disk graphs.

• Number of nodes n are chosen for the operations, where n=10,20,30,…,100.

• Field size (area) is 600x600 unit.• Transmission range of each node is 80

unit

Number of Clusters

Number of nodes

If the number of nodes increases, the ratio of the cluster with respect to the number of nodes decreases.

Relation between the number of nodes and clusters in node-move-in of case 1 and case 2.

0

5

10

15

20

25

node-move-in of case 1

node-move-in of case 2

0600

12001800240030003600420048005400600066007200780084009000

10 20 30 40 50 60 70 80 90 100

Number of Rounds

Number of nodes

Relation between the number of rounds of node-move-in in case 1 and case 2.

When network grows larger, the node-move-in operation in case 1 takes much more rounds than that of in case 2.

node-move-in of case 1

node-move-in of case 2

ConclusionConclusion Information

available to the nodes

Problem Completion rounds

Case 1 and Case 2 Broadcasting Case 1 node-move-in expec.

node-move-out

Case 2 node-move-in expec.

node-move-out

expec.

expec.

)(logqO

)(qO

|)(|TO

)log.|(| rTO

)( pO

subtree-T and new of nodesneighbor ofnumber

.G in subgraph completedisjoint the

ofnumber smallest the, than less clusters theofnumber

q

pp G

)|(| tpTO

Homework/Assignment

1. Given a graph G, describe the definitions for Independent Set (IS), Maximum Independent Set (MIS), Dominating Set (DS), Minimum Dominating Set (MDS), and disk graph, respectively.

2. Search for the distributed algorithms for finding MIS and MDS for a give graph G, respectively. What are the time complexity? 3. Given a disk graph and its MIS, design a centralized algorithm for forming CNET(G).

Reference:(1) [1](2) B.S. Chlebus, L. Ga¸sieniec, A.M. Gibbons, A. Pelc, and W. Rytter.Deterministic broadcasting in ad hoc radio networks. DistributedComputing 15, pages 27–38, 2002.