Lab Seminar 2009 06 17 Description Based Ad Hoc Networks

1

Description Based addressing and routingin

Cluster-based Ad-Hoc network for

Home Environment

By

Tharinda Nishantha Vidanagama*

Supervised by,

Prof. Hidenori Nakazato**Graduate school of Global Information &

Telecommunication Studies, Waseda University.

GITS Seminar

2

Ad Hoc Networks

The Ad Hoc peer network is created spontaneously as two participating nodes come within the reach of one another.

As the number of participating nodes and their positions vary the network it self reconfigures allowing all nodes to communicate even if they are not within their transmission range.

The main advantage of Ad Hoc networks is that they do not require any infrastructure to operate.

But the main disadvantages are Highly dynamic topology due to the mobility of the nodes, i.e. the links are not

stable and bandwidth is limited. Mobile devices are powered by batteries, putting an energy consumption

constraint

3

Ad Hoc Networks

Increasingly wireless technology is build in to many devices.

Networking these devices can bring many benefits.

Range of application is vast. Military Disaster Service extension.

4

Description based Ad Hoc Networks

Communication failures are considered normal in Ad-hoc networks.

Description based addressing instead of host based(IP) addressing is more appropriate because it will not rely on a single path, but uses multicasting .

In description-based routing, selective information dissemination can be implemented.

Easy to understanding and manage the network for home environment.

This will minimize control data among peers and facilitate group communication.

5

Ad Hoc Network at Home !

Increasingly new functions are built in to home appliances.

Home appliances fitted with wireless communication modules would become more common in near future.

The system for home networking provides information required for inter-working environment for home appliances.

The nodes in this environment could also have the following constraints, Low memory Low transmission distances Some nodes may have limited power supply.

6

Add Hoc home network - Approach

1. Cluster Based Ad Hoc Network

2. Description Based Routing and Addressing

7

1. Cluster Based Add Hoc Network Why?

At home most appliances are clustered in places. Kitchen Living room Bed room etc.

Clustering also reduces the amount of control data exchanged among the nodes.

Only a portion of the nodes are involved in the actual routing process.

8

Cluster Based Add Hoc Network Organization

Everyone knows its one hop neighbors (Neighbor Table).

Cluster heads know their neighboring cluster heads (Cluster Adjacency Table).

Member nodes send their messages to their respective cluster heads to be forwarded to their destinations.

Only cluster heads and gateways forward messages.

9

Cluster Based Add Hoc Network Organization…

Neighbor Table of M1

Head H1

M2 Member

G2 Gateway

Cluster Adjacency Table of H1

H2 G1

H3 G2

Cluster Adjacency Table of G2Neighbor Table of G2

Head H1

M1 Member

G3 Gateway

H1 H1

H3 G3

10

Problem of Routing Does a cluster head know about all other cluster heads?

This requires high memory capacity !

How to route messages beyond known cluster range?

A fixed size buffer is used to keep track of the nodes that are beyond range.

Rather than using a table structure this helps reduce the memory usage.

There is no need to store the full identities or routs.

11

We assume that identifiers given to the nodes in the network i.e. home appliances are similar to the following examples.

Kitchen oven, Kitchen television, Living room television etc.

In this natural language usage, a general location is identified from the beginning words and becomes more specific within that particular area by the next words.

Our routing algorithm routes the data messages in the direction of the destination.

When the message finally arrives at the particular region, there would be nodes that store the exact identifier in their Neighbor table, thus allowing the message to be delivered to the correct destination node.

2. Description Based Routing and Addressing

12

How the buffer is used?

If a cluster head or gateway node comes across another cluster head that is beyond its range, it inserts the name in to the buffer.

E.g: if the required destination is “kitchen oven” but the buffer entry may be “kitchen”.

This partial match is enough to ensure that the message gets routed in the direction of the kitchen.

If the entry is in the 1st position of the buffer it would be as follows.

Buffer Usage

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

nehctik

13

Buffer storage For the simulation purpose we use a fixed size character buffer. Elements are inserted in a way that some of the information about

old entries are overwritten by new entries. If

S - Buffer size p - Number of parts the buffer is divided in to P - The next position in the buffer to be occupied B - 1 if data written backwards (If entry number is and odd number) F - 1 if data is written forward (If entry number is an even number) f - Fibonacci series number.

When an entry is made in to the buffer the starting position (T) for an entry is given according to the following formula.

Starting Position T= ( / )*( )S p P B F

14

Buffer storage The number of characters allocated for an entry in the buffer is given

by the following formula,

Number of characters =

The actual character positions are given by a modified version of the Fibonacci series (0, 1, 2, 3, 5…), where we have omitted the repetition of number 1. Character position in the buffer is given by the following formula,

Character Position =

( / ) 1ps

; 1.; 1.{T f F

T f B

15

Description Based Routing

When a node receives a message, If destination is this node

Consume dataElse if I am Gateway/Cluster head and if this data has not already been

forwardedLook for destination in neighbor table

If not found Look for destination in Cluster Adjacency table

If not foundLook for partial match in Cluster Adjacency table

If not found Look for partial/complete match in the Buffer.

If foundForward message

ElseDiscard message.

16

A simulation program was developed in order to evaluate the above proposal.

We have used an environment, which is a flat surface 25*25 square units, where 100 nodes are statically located.

The size of the buffer was limited to 128 characters. Each node has various transmission distance (maximum of 3 units)

and priorities. Each node will send a number of messages to randomly selected

destinations.

Evaluation

17

It shows on average a 92.4% success rate on message delivery.

Evaluation

0

10

20

30

40

50

60

70

80

90

100

1 2 3 4 5 6 7 8

Percentage

Sent Data(/100)

Received Data (/100)

The number of messages (Y)

Simulation execution (X)

Success rate

18

It shows on average a 100% success rate on message delivery with the current environment setting.

But… Loss is high !!!.

Evaluation – After Simulation Upgrade.

Percentage (Y)

Data Message loss due to wrong routingSingle Buffer

187.4 179 183.93 180.9 185.6 190.33 191.23174.53

191.09

167.88

100 100 100 100 100 100 100 100 100 100

020406080

100120140160180200220

5 10 15 20 25 30 35 40 45 50

Data msgs/100

loss

success

19

The loss occurred because of the use of single buffer. Buffer entries do not store the direction of that particular entry. Therefore the

messages maybe forwarded to the wrong way as well.

Evaluation – The loss?

  GW

CH

CH

128

128

128

CH

128

128

CH

GW

Bidirectional link

Buffer

Cluster Head

Gateway

20

Routing nodes are given a buffer for each outward link. The nodes can now compute the next hop of the messages in the

right direction.

Solution – Directional buffers

GWCH

128

128

128

CH

GW

Bidirectional link

Buffer

Cluster Head

Gateway

21

The loss due to wrong routing is significantly reduced by the directional buffers.

Simulation – Directional buffersData Message loss due to wrong routing

Directional Buffer

26.434.6 34.8 35.05 36.44

30.4739.34

33.48 30.44

45.76

100 100 100 100 100 100 100 100 100 100

0102030405060708090

100110

5 10 15 20 25 30 35 40 45 50

Data msgs/100

loss

success

22

When a message is received by a node that has many outward links, it checks each buffer for a match with the destination.

If more than one buffer is positively matched the node has to broadcast the message to all outward links.

Thereafter only the next hop nodes who know the destination will forward the message, and the others will register a lost message.

Directional buffers – remaining loss

23

The loss due to wrong routing is significantly reduced by the directional buffers.

Simulation ComparisonData Message loss Comparison

at 100% success rate

1157.8 1141.1 1145 1139.2 1151.6 1158.93 1146.37 1142.6 1162.42 1159.44

187.4 179 183.93 180.9 185.6 190.33 191.23 174.53 191.09 167.88

26.4 34.6 34.8 35.05 36.44 30.47 39.34 33.48 30.44 45.760100200300400500600700800900

1000110012001300

5 10 15 20 25 30 35 40 45 50Data msgs/100

Broadcasting

Single buffer

Directional buffer%

24

Considering that a home environment does not contain large number of nodes, this algorithm can be effectively implemented.

The use of clustering has further supported in reducing the network traffic and routing overhead.

Use of descriptions improves the user understandability of the system.

Use of directional buffers will significantly reduce the message loss and improve resource usage.

Conclusion

25

We will study the performance and issues when movement is introduced to the network nodes. In the home environment elements such as cell phones and laptops etc, will have movement associated with them.

We would also like to find the optimal buffer size for a given number of nodes.

Divide the same buffer proportional to the number of entries in each direction.

Further decrease the amount of control information held and transmitted by nodes.

Future extensions

26

Questions ?

27

Thank you !