22
Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

Embed Size (px)

Citation preview

Page 1: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

Infocom’07Authors:Liqian Luo, Chengdu Huang, Tarek

Abdelzaher John Stankovic

Presented By Rohini KurkalUnder Guidance of Dr.Bin Tang

Page 2: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

Contents IntroductionSystem DesignImplementationPerformance EvaluationPossible Extensions

Page 3: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

IntroductionApplications: -Environmental monitoring/data logging

No real time communication b/w sensor and sink Disconnected system

EnviroStore - cooperative storage system for sensor networks suitable for disconnected operationIt is storage-centric rather than communication-centricMajor concern -maximizing effective storage capacityImplemented in nesC for TinyOS and evaluated in TOSSIM

Page 4: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

Storage centric paradigmShould be simple and lightweight

Micaz has only an 8MHz 8-bit processor and a 4KB RAM,

Iris has 8KB RAM, 512KB Measurement Flash Memory

Sensor nodes do not maintain files but just writes data to collection station and never read the data they write

Data redistribution–must improve overall storage utilization

Page 5: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

System Design

Sink – process that runs on a user’s PC, identified by a regular IP address and a well known TCP port

Data mule-collects data wirelessly from encountered nodes and dumps these data later to the base station

Two types of data mules: - Intentionally relay data b/w the sink and the

sensor nodes - Opportunistic data upload

Page 6: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

System Model

Page 7: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

System DesignData redistribution is used to maximize the effective storage space of the sensor networkIn-network Data Redistribution:

Sensor nodes are in a single network Overloaded nodes offload data to neighboring

empty nodes

Cross-partition Data Redistribution: Disconnected network Overloaded nodes upload to data mules Data Mules offload to under-loaded nodes

Page 8: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

In-network data redistributionUses lazy-offload scheme to save energy – postpones

data balancing until the storage overflows

Overloaded nodes should satisfy below conditions: Ri < RTH and Ri

’ - Ri > Rimbalance

Where ,

Ri = Remaining Storage size

RTH = Threshold to delay data transfer

Rimbalance = Parameter to allow local imbalances

Ri’ = Average remaining storage

Page 9: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

Contd..Bad Idea:

Selecting neighbor with largest remaining free space

This can cause data ping-pongPrevent data ping pong - bound the amount of

data transferRemaining storage & remaining node energy

must be checkedNode should not invoke or accept data

redistribution unless its estimated energy lifetime > estimated storage lifetime

Page 10: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

Cross-partition data redistribution

Uses data mules Discriminate nodes (overloaded and under-

loaded): Data mule calculates its free storage value R’m as

the weighted sum αR’+(1-α)Rm.

α = 1 : mule favors redistribution to neighborhood

α =0 : emphasizes uploadConserve power and reduce message collision,

nodes use back-off timers

Page 11: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

Transition state of sensor node

Page 12: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

System Architecture of sensor nodes

Page 13: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

ImplementationImplemented using nesC in TinyosLocal storage space of nodes is organized into a

circular buffersUses:

It consumes minimum code and data memory organizes space as continuous data chunk eliminates the need for free space management

mechanisms prolongs flash lifetime by balancing write access to

different locations

Page 14: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

Log FilesLog-array file :

Simultaneously written by different nodes Generates a sequence of log items

Logs attributes of environmental events independently monitored by multiple nodes

Log-sequence file:one writer at a time Multiple nodes must coordinate with each otherMaintains unique & continuous serial numbersUsed in EnviroSuite

Page 15: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

Log-Array File

Page 16: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

Log-Sequence File

Page 17: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

PerformanceEvaluation Basic deployment configuration: field of 80 ×

80 ft2 , 36 nodes were deployedTwo Scenarios:Scenarios 1: Single Disconnected sensor network Scenario 2 :Partitioned sensor network with data mules

Page 18: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang
Page 19: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

Comparison of data storing rate at different time

Page 20: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

Why EnviroStore is different ?Used for disconnected sensor networksExtra constraint of limited energy – use lazy

offloadResource limitation of individual nodes Load balancing must be dependent only on

local informationHas additional challenge of redistributing

data between entities that are disconnected

Page 21: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

Possible Extensionsuse of controllable data mules to optimize

data redistribution and uploaddata replacement policies to maximize the

total amount of information instead of just the amount of stored data

performance evaluation of EnviroStore on real hardware platform

Page 22: Infocom’07 Authors:Liqian Luo, Chengdu Huang, Tarek Abdelzaher John Stankovic Presented By Rohini Kurkal Under Guidance of Dr.Bin Tang

THANK YOU