Dissemination protocols for large sensor networksFan Ye, Haiyun Luo, Songwu Lu and Lixia ZhangDepartment of Computer ScienceUCLA

Chien Kang Wu

Outline

Introduction Current Dissemination strategy

Reverse path forwardingCost field based forwardingRouting with virtual hierarchyAdditional approaches

Future work

Introduction(1/2)

Goal : Communicate organized data using data-centric paradigmScalable and distributed solutionEnergy efficiencyRobustness

DefinitionData source : sensor node which generate dataSink : user collect data from sensor networks

Introduction(2/2)

Assumption Application semanticsLocation awarenessStationary nodesDense deployment

Solution roadmapQuery-Reply process Install dissemination states in intermediate

nodes

Outline

Introduction Current Dissemination strategy

Reverse path forwardingCost field based forwardingRouting with virtual hierarchyAdditional approaches

Future work

Reverse path forwarding

Including: Declarative routing protocol (DRP) Directed Diffusion (DD)

Sink sends out a query using flooding strategy through the network

Data flow in the reverse direction of query Set up forwarding state in the form of vector

Sensor broadcast data to neighbors,

and neighbor recursively forward it .

Declarative routing protocol

Establish a routing tree for every sink Use factors to select which node of the next

Neighborhood the query should be sent Using cashed data to improve efficiency

Reachability ,remaining energy directionality, link quality

Declarative routing protocol

Base on :Location awareness Node with small buffer

Faced challenges:Buffer managementRoute inconsistencyData aggregation strategy

Directed diffusion

Similar to DRP, but focus more on scaling to multiple sinks

Node do not keep per sink state Each node has a cache to detect loop and

drop redundant packets sink can use reinforcement mechanism to

help neighbors select the best quality path To handle network dynamics , source need

to maintain alternative paths

Directed diffusion

Base on:Location awareness Node with small buffer

Faced challenges:Maintain alternative paths to handle nodes failureQuick Path repairing methods In-network processing

Problems in reversed path

Overhead has to be paid to maintain

vector states Sink mobility problems

Outline

Introduction Current Dissemination strategy

Reverse path forwardingCost field based forwardingRouting with virtual hierarchyAdditional approaches

Future work

Cost-field based forwarding

Each node store forwarding state named scalar denoting the node’s distance to sink

Scalar of all nodes forms a cost-field Cost-field is per-sink based, a node keep

cost for each sink Data report flow from higher cost to

smaller cost

Hop count, energy consumption physical distance

Cost-field based forwarding If query-packet(ADV) flow from node i to node j ,

C(i new) = C(j) + C(i , j) (set 0 at sink)

set C(i) to C(i new) , if C(i) > C(i new) There is no loop in cost-field Two-ways to make

forwarding decision Receiver-decided Sender-appointed

Receiver-decided

Data sender includes its cost in a report and broadcast to every neighbor

Let receiver decides forwarding or notOnly receiver with cost less than sender may

forward the report Robust data forwarding:

Typically , several neighbors with smaller cost than the sender

Need strategy to prevent data redundancy problem

Sender-appointed

Single path forwarding, sender choose only one neighbor for each report

Can not ensure robustness, need to maintain states regarding which neighbors are still alive

Statistically distribute strategy ,save more energy than receiver-based

Problems in Cost-field based

Mobile sinks problems Does not scale well with numerous sinks,

each sink needs a separate cost at every node

Outline

Introduction Current Dissemination strategy

Reverse path forwardingCost field based forwardingRouting with virtual hierarchyAdditional approaches

Future work

Virtual hierarchy

Sensor nodes have different functionality,

network hierarchy is formed during data dissemination

Sink mobility support Including:

Two-tier data disseminationLow-energy adaptive clustering

Two-tier data dissemination

Data source construct virtual grid infrastructure for query and data forwarding, each grid is an α*α square

Data Source propagate data announcement to reach all other crossing of grid, called dissemination point

Sink forward query to its upstream dissemination node dissemination node further forward the query toward data source

Grid infrastructure

Two-tier query-reply process

Low-energy adaptive clustering

Designed towards energy-optimal network organization

Assumption : Sensor can adjust power consumption, adapt network

topology Sensor’s max power is able to communicate with the sink

Data message from sensor first transmitted to local cluster head and then forward to base station

System using distributed algorithm to elect a number of cluster heads

Virtual hierarchy summary

Advantage:Structure suitable for Data aggregation Mobile sink is feasible

Faced challenges:Structure maintain problemCluster head election problem

Outline

Introduction Current Dissemination strategy

Reverse path forwardingCost field based forwardingRouting with virtual hierarchyAdditional approaches

Future work

Additional approaches

Real-time deliveryDefine: distance d ,packet lifetime t desired velocity v = d / tRequired velocity is updated at each hopA node can use multiple FIFO queue with

different priority to handle packetsChoose neighbor with high velocity to forward

the report

Outline

Introduction Current Dissemination strategy

Reverse path forwardingCost field based forwardingRouting with virtual hierarchyAdditional approaches

Future work

Future work

Network topology:Location awareness Hierarchy structure maintain Path maintain issueAlternative paths reinforcement

General issues:Buffer management and data aggregationData redundancy problem