lecture_03_3_wireless sensor network protocols & algorithms

Wireless Sensor Network Protocols• Primary theme: building • Data centric routing In-network: Application processing, Aggregation, Query processing Adaptive topology, Geo-Routing MAC,

Sensor Network Protocols

By

Quan Le-Trung, Dr.techn., Summer.2015

http://sites.google.com/site/quanletrung/

Wireless Sensor Network Protocols

• Primary theme: building

• Data centric routing

In-network: Application processing,

Aggregation, Query processing

Adaptive topology, Geo-Routing

MAC, Time, Location

User Queries, External Database

Data dissemination, storage, caching

Motivations: An Integration View of Wireless

Sensor Networks

SensorNets Database

+Local Storage +External Storage +Distributed Storage +Data-Centric

+Multi-Resolution +Predictive

With Query:

+Query Dissemination

Data Dissemination Data Aggregation

Data-centric Routing

based Routing

Location-Energy-efficient Routing Flooding,

Gossiping

Hierarchical Routing

Network-Flow Routing

Data/Query Dissemination &

Different Routing Mechanism

(Routing & Network Infrastructure)

Query Dissemination Multi-Query Optimization

Data Storage Model

Sensor Network Architecture

Management Station:

+Queries delivered by Web Server from Monitoring User(s) are interpreted, summarized, and optimized here, before injecting into WSNs through Sink/Gateway

+Context reasoning based on External DATA/CODE/CIA Repositories

+Can be implemented as a component on the Sink/Gateway

Premier Network, e.g., Ethernet 802.3, WLAN 802.11

External DATA/

CODE/CIA

Repositories

Web Server Management Station

Sink/Gateway

Local DATA/CODE/

CIA Repositories

(1)/(6) (2)/(5)

(3b)/(4b) (3a)/(4a)

(3b)/(4b)

(3 b)/ (4 b)

Motivations: An Integration ARCH

for Wireless Sensor Networks

Logical Network Infrastructure Information Storage & Processing

Data Dissemination ARCH

Information Reduction Techniques

+Different sub-classes in Fig 2 +Trade-off Energy-efficiency vs

+Data accuracy & freshness +Latency

DATA Storage ARCH.

+Standard sink model

+External storage

+Local storage

+Distributed indexing & storage

+Distributed data-centric storage

(DCS)

+Predictive storage

+Multi-resolution data storage

QUERY Processing Mechanism (DATA Mining)

+Push vs Pull vs Hybrid +Single- vs Multi-Queries +Value-range vs Location-range Queries +Structured vs Unstructured Queries +Conditional vs Traditional Queries +Single- vs Multi-Dimensional Queries +DHT, GHT, DIMENSIONS

+DIFS, DIM

ATTRIBUTE

+Distributed attribute allocation storage (DCS- based)

Tier-1 Tier-2 Tier-3

Information Reduction Techniques

– Connected Correlation Dominating Set [CCDS]

– Clustered Aggregation [CAG]

– Sparse Aggregation [SAG]

– Self-based Regression [SBR]

– Bayesian Inteference [Infer]

– Temporal Coherency-Aware in-Network Aggregation [TiNA]

• TAG-based Aggregation

– SUM/COUNT/AVG/MAX/MIN

– Q-Digest [MEDIAN], Range Query

• Accuracy

Information Reduction Techniques

Index Objectives Target Timing Architecture Reduction Technique Coding Energy-efficiency Correlated data Yes Tree Compression

Wavelet Energy/Storage-efficiency Spatio-temporal query Yes Quad-Tree Compression

CWS Energy-efficiency/Low-latency Structural regularity data Yes Tree Compression

LR Energy-efficiency/Accuracy Correlated spatial data No Clustering Compression/Accuracy

SRA Energy-efficiency/Accuracy Feature extraction query Yes Clustering Compression/Accuracy

Infer Energy-efficiency/Accuracy Unobserved data Yes Flat Selection

q-digest Energy-efficiency/Bandwidth Quantile/Range query No Tree Aggregation

apx-MEDIAN Energy/Storage-efficiency Quantile query No Tree Aggregation

Quan Le-Trung, Paal E Engelstad , Tor Skeie , Amirhosein Taherkordi , and Hai N Pham , (2009), “Information Storage, Reduction,

In-Network Processing

The Key to Sensor Network scalability and Realization

• Gupta and Kumar pointed out fundamental limits of

large scale wireless networks (per node throughput O(1/sqrtN)

• However, S Servetto shows that result holds only for

independent nodes (Mobicom 2002)

– Densely deployed sensor network data will be

correlated and can be aggregated

• Scalability and lifetime will depend on techniques

for in-network processing of data

Accommodating in-network Aggregation

Explores aggregation techniques that are

Tricks of the Trade …

• How do you ensure an aggregate is correct?

– Compute it multiple times

• How do you reduce the message overhead of redistributing queries?

– Piggy back the query along with data

• Nodes can take advantage of multiple

parents for redundancy reasons

Supporting Aggregate Queries Over Ad-Hoc Wireless Sensor

Motivation: Sensor Nets and In-Network Query Processing

Oriented

Mechanism

– Not subject to Moore’s law!

The Tiny Aggregation (TAG) Approach

network

local sensor data and data from children

– Aggregate local and child data

optimizations can be applied

SQL Primer

– Some extensions clearly necessary, e.g for sample rates

– One column for each reading-type, or attribute

– One row for each externalized value

 May represent an aggregation of several individual readings

SELECT {agg n (attr n ), attrs}

Aggregation Functions

Agg n ={f merge , f init , f evaluate }

F merge {<a 1 >,<a 2 >}  <a 12 >

f init {a 0 }  <a 0 >

F evaluate {<a 1 >}  aggregate value

(Merge associative, commutative!)

Example: Average

AVG merge {<S 1 , C 1 >, <S 2 , C 2 >}  < S 1 + S 2 , C 1 + C 2 >

Partial Aggregate

Query Propagation

 Deliver the query to all sensors

 Provide all sensors with one or more duplicate free routes to some root

flooding approach

– Query introduced at a root;

rebroadcast by all sensors until it reaches leaves

– Sensors pick parent and level when they hear query

6

P:1, L:2 P:1, L:2

P:3, L:3 P:2, L:3

P: parent L: level

Illustration: Pipelined Aggregation

Illustration: Pipelined Aggregation

11

FROM sensors

Illustration: Pipelined Aggregation

21

FROM sensors

Illustration: Pipelined Aggregation

31

FROM sensors

Illustration: Pipelined Aggregation

31

FROM sensors

Illustration: Pipelined Aggregation

31

FROM sensors

most d-1 epochs

epochs, except during small communication window

1

4

5

readings on each epoch

– If it belongs to a stored group, merge with existing record for that group

– If not, just store it

Example: SELECT max(light) FROM sensors

Logical Network Infrastructures

• Data Dissemination ARCH.

• Grid-based [Two-Tier Data Dissemination (TTDD)]

• Query/Data Dissemination ARCH.

– Tiny Aggregation [TAG]-based Tree

– Extensions of TAG-based Tree

• Approximation, Synopsis Diffusion

– Tributary-Delta

– Sweep

Cluster Architecture

Chain Architecture

Young-Long Chen, and Jia-Sheng Lin, “Energy efficiency analysis of a chain-based scheme via

intra-grid for wireless sensor networks,” Elsevier Computer Communications, Volume 35, Issue 4, 15 February

Tree Architecture

http://www.ece.gatech.edu/research/labs/bwn /WMSN/Images/Testbed/testbed.jpg

http://www.intechopen.com/source/html/37868/

media/fig9.jpg

Grid Architecture

http://ars.els-cdn.com/content/image/1-s2.0-S0140366412003106-gr1.jpg

Tributary-Delta Architecture

cdn.com/content/image/1-s2.0- S0037073802002567-gr7.jpg

http://ars.els-Logical Network Infrastructures

Index Fault-tolerance ARCH Underlying protocol Target Sink Applications

LEACH Rotating cluster

head

Clustering Push

CSMA MAC, different CDMA codes in each cluster DATA Fixed

Constant monitoring Periodic data reporting HEED Rotating cluster

head

Clustering Push

Ad-hoc routing for cluster communication DATA Fixed Environmental monitoringPEGASIS Chain

inter-reconstruction

Chain Push

Location-aware CDMA/non-CDMA DATA Fixed Environmental monitoring

TTDD Upstream info

duplication, timeout

Grid Push/Pull

Location-aware Geographic forwarding

DATA QUERY Mobile Event of interest

EADAT Rotating branch

points

Tree Push

Transceiver with on/off radio capability DATA Fixed

Monitoring abnormal events

TAG Multi-path

routing

Tree Push/Pull Ad-hoc routing

DATA QUERY Fixed

Monitoring and data collection tasks

Synopsis

diffusion

Multi-path routing, order- &

duplicate insensitive synopses (ODI)

Ring Push/Pull

Broadcast wireless communication

DATA QUERY Fixed

Aggregation: Sum, Count, Avg, Medium, Uniform, Max, Min

Approximation

Multi-path routing, duplicate- insensitive sketches

Ring Push/Pull N/A

DATA QUERY Fixed

Aggregation: Sum, Count, Avg, Max, Min

Tributary-Delta Multi-path routing

Tree-based

Combining Tree &

Multi-path routing N/A

DATA QUERY Fixed

Aggregation: Sum, Count, Avg, Medium, Uniform, Max, Min, Quantiles

Sweep

One-to-many downstream neighbors, local Wavefront CSMA MAC

DATA QUERY Fixed

Aggregation: Sum, Count, Avg, Max, Min

Why databases?

• Sensor networks are capable of producing massive amounts of data

– Sensor networks should be able to

• Accept queries for data

• Respond with results– Users will need

• An abstraction that guarantees reliable results

• Largely autonomous, long lived network

• Efficient organization of nodes and data will extend

network lifetime

– Database techniques already exist for efficient

data storage and access

Differences between databases and sensor networks

– Multi-hop network– No global knowledge about the network

– Frequent node failure

– Energy is the scarce– Resource, limited memory

– Autonomous

Bridging the Gap

• What is needed to be able to treat a sensor network like a database?

– How should sensors be modeled?

– How should queries be formulated?

Traditional Approach: Warehousing

• Data is extracted from sensors and stored on a front-end server

• Query processing takes place on the front-end.

Warehouse

Front-end

Sensor Nodes

What We’d Like to Do:

Sensor Database System

• Sensor Database System supports distributed query processing

over a sensor network

Sensor DB Sensor

DB

Sensor DB

Sensor

DB

Sensor DB

Sensor DB

Sensor DB

Front-end

– Time between data is generated on

sensors and answer is returned

• Limited resource usage

– Energy consumption

Data-Centric Storage in

Sensornets

Sylvia Ratnasamy, Scott Shenker,

Brad Karp, Ramesh Govindan, Deborah Estrin

ICSI/UCB/USC/UCLA

 Sensornet

number of small sensing devices equipped with

• processor • memory • radio

 Data Dissemination Algorithm

 Observation

Images of intruders detected

Existing Schemes

 External Storage (ES)

 Local Storage (LS)

 Data-Centric Storage (DCS)

External Storage (ES)

Local Storage (LS)

Event Data Event

Data

Local Storage (LS)

Event Data Event

Data

Data-Centric Storage (DCS)

 Events are named with keys

 DCS provides (key, value) pair

 DCS supports two operations:

♦ Put (k, v) stores v ( the observed data ) according to

the key k, the name of the data

♦ Get (k) retrieves whatever value is stored

associated with key k

 Hash function

to the same location

DCS – Example – contd

PDA

elephant

fire

Geographic Hash Table (GHT)

 Builds on

♦ Peer-to-peer Lookup Systems

♦ Greedy Perimeter Stateless Routing

GHT

GPSRPeer-to-peer

lookup system

 Not robust enough

♦ Nodes could move (new home node?)

♦ Home nodes could fail

 Perimeter Refresh Protocol

♦ Extension for robustness

♦ Handles nodes failure and topology change

 Structured Replication

♦ Extension for scalability

♦ Load balance

Comparison Study - contd

O(n 1/2 ) for point-to-point routing

Comparison Study -contd

 D total, the total number of events detected

 Q , the number of event types queries for

are Q queries in total.

access point

Comparison Study – contd

D n

Q  total 

) (

2 Q summary

within the sensornets

key-value pair is stored at a node in the vicinity of the

location to which its key hashes

Perimeter Refresh Protocol (PRP) and Structured

Replication (SR)

sensornet

DIMENSIONS TM : A Multi-Resolution Storage

Architectures for Sensor Networks (Ganesan et al)

increasing spatial resolution

increasing temporal resolution

 Per-Node Temporal Data Processing

 Progressive encoding : more lossy storage of older data

 Design Principle : System guarantees lossless multi-resolution data collection within time T of event occurrence Older data is stored ONLY for long-term query processing, which can tolerate greater loss

 Spatial Data Processing

 Nodes in the network form a logical spatial hiearchy

 At each step of the hierarchy, further summarization of the data takes place.

Routing & Physical Network Infrastructures

• Low Energy Adaptive Clustering Hierarchy [LEACH]

• Geographical and Energy-Aware Routing [GEAR]

• Greedy Perimeteer Stateless Routing [GPSR]

Routing Protocols in WSNs

Directed Diffusion for Sensor

Networks

Chalermek Intanagonwiwat,

Ramesh Govindan, Deborah Estrin

Presented by: Prince Samar

Data Naming

 Task are named: Attribute – value pair

 Selecting naming scheme important

 No globally unique ID for nodes: only locally unique

rect = [-100,100,200,200]

timestamp = 01:20:40

Interest & Gradient

 Interest describes a task needed to be done by the sensor network

 Interests are injected into the network at sink

 Sink broadcasts the interest

 Interval specifies an event data rate

 Initially, requested interval much larger than needed

 Node maintains an interest cache

 Specifies a data rate and a direction (neighbor)

Data flows from the source to the sink along the gradient

Interest Propagation

 Flooding

 Constrained or Directional flooding based on location

 Directional Propagation based on previously cached data

Source

Sink

Interest Gradient

Data Propagation

 Reinforcement to single path delivery

 Multipath delivery with probabilistic forwarding

 Multipath delivery with selective quality along different paths

 Reinforce one of the neighbor after receiving initial data

 Neighbor(s) from whom new events received

 Neighbor who’s consistently performing better than others

 Neighbor from whom most events received

Reinforcement

Performance Evaluation

 Simulator: ns-2

 Network Size: 50-250 Nodes

 Transmission Range: 40m

 MAC: Modified Contention-based MAC

 Transceiver Energy Model: mimics a “sensor radio”

 660 mW in transmission, 395 mW in reception, and 35 mW in idle

 Comparison with

 Flooding

 Omniscient multicast

Average Dissipated Energy

(Sensor radio energy model)

0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018

Impact of In-network Processing

0 0.005 0.01 0.015 0.02 0.025

Impact of Negative Reinforcement

0 0.002

0.004

0.006

0.008

0.01 0.012

Average Dissipated Energy

(802.11 energy model)

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14

 Application-level data dissemination has the potential to improve energy efficiency significantly

 Data-centric dissemination

 Reinforcement based adaptation of paths

 Duplicate suppression and aggregation

 MAC for sensor networks needs to be designed carefully