Load Balancing in Sensors Network

8/6/2019 Load Balancing in Sensors Network

1/23

Effective Routing Scheme to Perform

Load Balancing

in Grid Wireless Sensor Network

Presented By:

Jaya ChughM.Tech CSE

DCRUST, Murthal

DEENBANDHU CHHOTU RAM UNIVERSITY OF SCIENCE & TECHNOLOGY


2/23

Index

Introduction

Existing Routing Scheme Problem Identification

Proposed Scheme

Conclusion


3/23

Wireless sensor networks

Wireless sensor network (WSN) consists of scattered sensors

which monitor physical or environmental conditions such as

temperature, sound, pressure, motion or pollutants. Thesetiny sensor nodes communicate with each other using low

power wireless data routing protocols. Sensor network

routing protocols and algorithms must possess self-

organizing capabilities .


4/23

Sensor Networks Communication

Architecture

Fig1 :Architecture of WSN .


5/23

Components of a sensor node

A sensor node is made up of four basic components:

sensing unit, processing unit, transceiver unit

power unit[6].

Fig2:components of sensornode


6/23

Intro

ductio

n.

In this paper we design load balancing algorithm for grid wirelesssensor network. In such type of network, the load increases on somecenter node in the network. We assume an All to All communicationmode in an N N grid sensor network. We explore routing algorithms

which balances the load on center node. The problem is divided intotwo scenarios, one being the static case (without failed nodes), and theother being the dynamic case (with failed nodes). In static Network weconsidered load balancing over all the nodes in the grid to be thecriterion for deciding the ecency of the routing algorithm using H-Vrouting. In dynamic network scenario, Success Ratio is taken in whichsuccessful transmission of message from source to destination is more

essential than load balancing over the nodes in the network. Theprimary focus of this research is to develop into deterministic routingstrategies which perform better load balancing of the mesh in an all toall communication scenarios, using shortest paths.


7/23

Routing Protocols

Low Energy Adaptive Clustering Hierarchy (LEACH)

Power Efficient Gathering in Sensor Information Systems(PEGASIS)

Sensor Protocol for Information via Negotiation (SPIN):

Fig3 SPIN protocol

Destination-Sequenced Distance Vector routing (DSDV):


8/23

Existing Routing Scheme

Flooding

Gossiping

Data Centric Routing

Fig 4:Gossip basedApproach

Fig5 Illustration of AC versus DC routing


9/23

Contd

. Directed Diffusion

Rumor Routing

Servetto routing algorithm which reduces the load in NxNgrid network on the central node in a single sourcesingledestination communication. It divides the network intoexpansion and compression phases. Sensor nodes belong todierent diagonals of the grid. During expansion phase,the load per node decreases with the increase in number of

nodes on diagonal. During the compression phase, thereverse process is carried out and with the decrease innumber of nodes on each diagonal, the load per nodeincreases. All the routing algorithms mentioned do notaddress the protocol performance in All to Allcommunication mode.


10/23

Problem Identification

Network congestion occurs when load exceeds the available capacity atany point in the network.It causes channel quality to degrade and lossrates to rise. Random deployments might result irregular topologieswhich in turn aect the routing strategy. The preference of regular

over an irregular topology simplies the analysis phase. Regular meshtopology for our problem is to be consider. In these networks, messagesare transmitted using multi-hop communications. Sensors performboth data sending and data routing. The central node will be heavilyutilized in routing and forwarding messages, while the corner nodesare less utilized. This uneven load distribution results in heavily loadednodes to discharge faster when compared to others. This causes few

over-utilized nodes which fail and result in formation of holes innetwork, resulting in increase of failed messages in the network. Arouting strategy developed should be such that it load balances thenetwork and prevents the formation of holes. Servetto proposed arouting algorithm ,reduces the load over the central node and increasesthe load over the corner nodes in a single sourcesingle destinationcommunication.


11/23

Contd

.

Fig6 Grid

wirele

ss Sen

sor Network


12/23

Keypoints

A static Network is a network with no node failures.

A dynamic Network is a network with failed nodes.

All to All communication phase is a mode in which all the nodes in thenetwork send and receive messages from all the other nodes in thenetwork.

Success-ratio is dened as the fraction of messages that reach theirdestination under node failure under shortest path routing

The primary focus of this research is to develop into deterministicrouting strategies which perform better load balancing of the mesh in

an all to all communication scenarios, using shortest paths thanexisting strategies.


13/23

Propos

edSch

eme

To avoid the congestion in case of center node we are representing thisresearch work. The basic idea is to share the load among thesurrounding nodes to the sensor nodes. We consider static scenarionetwork without node failures. One method using static network is

Horizontal Vertical (H-V), have been explored and compared withServettos method . These strategies are applied in an All to Allcommunication modes on an N N grid network and their resultantload distributions are analyzed. The performance of load balance isjudged based on the maximum and minimum loads of the grid whendifferent routing strategies are applied. Figure 7 shows the gridstructure considered.


14/23

Contd

..

Fig7 Grid Topology


15/23

Assumptions:

We are having a Grid Topolgy for wireless sensor network.

One end contains the transmitter and other contains transrecivers.

We consider the all to all communication mode.

The basic algorithm used for routing is on basis of shortest pathsearch.

There can be clustered approach.

There are innite-size buffers at each node of the grid to support theincoming and outgoing message packets.

Once the all to all communication phase begins, no new node during

the communication phase. In H-V described methods, consider a node (i, j) sending

message to node(k, l) and the simulations are performed on a10 10 grid. The dierent variants of horizontal-verticalmethod simulated:


16/23

Techniques for distributing the Load

1) Center Elimination: According to this method we will

remove the center node that has maximum load on it, now

instead of passing data from one center node it will move

from surrounding multiple nodes and instead of creating

on high hill on center it will distribute the load in form of

smaller hills surrounding to the center node.

Fig8 Center Elimination


17/23

Contd

.2) Avoiding Center: In this strategy the row distance (i x) and the

column distance (j y) of the node (i, j) from the center (x, y) iscalculated and the path having more distance from the center is taken.If the row distance is larger than or equal to column distance, then the

horizontal path is taken, otherwise the vertical path is taken.3) Actual Distance from Center: The actual distance from the center is

calculated using the co-ordinates of the nodes. At every step, there aretwo choices for the selection of the next node on the path to destination.The actual distance of these two nodes from the center is calculatedand the node which is more distant from the center is chosen as thenext node on the path to its destination. Repeatedly, this sameprocedure is applied at every node until it reaches its destination.


18/23

Contd

4) Average Distance from Center: In this strategy, all the nodes on thepath to the destination are taken and their average distance from thecenter, along the two available paths from source to destination iscalculated separately. Then the path having larger average distance

from the center is selected for routing the message from source todestination.

5) Minimum Distance from Center: In this method, all the nodes on thepath from source to destination are considered and their distancesfrom the center are measured. Along the two possible paths from thesource to destination, the minimum distance of the considered nodes iscalculated separately. Then the message is routed through the pathwhich has larger minimum distance of the two paths considered.


19/23

Contd

Fig9 Load Distribution On CornerNodes


20/23

Proposed Algorithm

Path (x, y) /* x, y are the source and destination nodes respectively*/

{

Find all the intermediate nodes between x and y

called nodes x1, x2 xn.

If (x (i) = CURRENT NODE AND x (i+1)

=NEXT NODEthen

{

if (Load x(i+1)) < THRESHOLD_VALUE)

{

x (i) = x (i+1); //Move to next node

}

else

{

Find Surrounding Nodes of x (i+1) and in

Queue call COMPROMIZING_NODES

Z=MIN_LOAD (COMPROMIZING_NODES)

//Find Minimum Load Node from the list of

compromising nodes and set it as next successful

node

set x (i) =z

}


21/23

Proposed Flowchart


22/23

Conclusion

In this paper, we studied shortest path routing algorithms namely-Horizontal-Vertical. We analyzed Servetto method and showed that H-V method performs better load balancing of the grid network thanServetto strategy, in an All to All communication mode. We will first

identify the center node for the exceeding load limit and then releasethis node from the communication path. Now its load gets distributedon the surrounding other nodes. This distribution will be taken onbasis of average distance from the center. After this central node isnow not overloaded. Holes would also minimize in this approach.Energy consumption on central node will be minimum so that thenetwork can be used more efficiently.


23/23

Load Balancing in Sensors Network

Documents

Transcript of Load Balancing in Sensors Network