SPIN: Sensor Protocols for Information via Negotiation

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL1

The X – Matrix Team

Adaptive Protocols for

Information Dissemination in

Wireless Sensor Networks

http://www.cs.ucl.ac.uk/students/fshariff/projects/spin

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL2

Who, What and HowThe X-Matrix Team- Wasif, Fahd, Philip, Muhammad and KumardevThe paper - Negotiation-based Protocols for Disseminating Information in Wireless Sensor Networks

byJoanna Kulik,Wendi Rabiner Heinzelman,and Hari Balakrishnan, Massachusetts Institute of Technology, Cambridge, MA, USA

The broad concepts outlined in the paper

Our Approach De-construction and Analysis of work Presentation Structure and Flow

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL3

Fundamental ConceptsWireless Sensor Networks

Sensors – typical size, weight, power characteristics

Sensor Networks are a subset of Ad Hoc Networks

Fixed / Mobile

Routing in Ad Hoc / Sensor Networks Traditional protocols – Classic flooding,

Gossiping Adaptive protocols – SPIN, Others

What are these so-called ‘adaptive protocols’?

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL4

Classic Flooding

B C

D

A

Sink Node

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL5

Problems with Classic Flooding

ImplosionA

B CD

(a)

(a)

(a)

(a)A B

C (r,s)(q,r)

q srData overlap

Energy Conservation

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL6

GossipingAlternative to Classic FloodingRandomisation to conserve energyAvoids implosion

B

C

D

A

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL7

The Ideal Protocol“Ideal” Shortest-path routes No wasted energy No redundant data

B

D E

FG

CA

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL8

SPIN: Negotiation and Dissemination

Overview of SPINApplication-Level ControlMeta-Data NegotiationSpin Messages ADV – New data advertisement REQ – Request for data DATA – The actual data

messageSPIN Resource Management

A B

A B

A B

ADV

REQ

DATA

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL9

SPIN family of protocolsPoint-to-Point SPIN-PP: a 3-stage handshake protocol for

point-to-point media SPIN-EC: SPIN-PP with a low-energy threshold

Broadcast SPIN-BC: a 3-stage handshake protocol for

broadcast media SPIN-RL: SPIN-BC for lossy networks

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL10

SPIN-PP

A

B

CE

D

DATA messageADV messageREQ message

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL11

SPIN-ECSPIN-PP with simple energy conservation heuristicWhen the low-energy threshold is observed, the node reduces its participation in the protocolNode can still receiveData messages cannot be transmitted

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL12

Questioning SPIN for Point-to-PointWhy use PP when we already have BC?Do we need energy conservation or is it application dependent?

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL13

Point-to-Point Media Simulations

Compare SPIN-PP and SPIN-EC with classic flooding, gossiping and the ideal protocol

Parameters of interest include: Data throughput Energy usage

Enhanced ns simulator

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL14

Simulation Testbed25 nodes, 59 edges25 data items3 items/node overlapAntenna reach: 10 m

No network losses or queuing delays

DataMeta-data

500 bytes

16 bytes

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL15

Unlimited Energy Simulations

Flooding fastest

-- SPIN-PP-- Ideal-- Flooding

SPIN-PP uses 3.5x less energy than flooding

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL16

Limited Energy Simulations

SPIN-EC distributes nearly the same amount as the ideal

SPIN uses energy at a much slower rate

-- SPIN-PP-- SPIN-EC-- Ideal-- Flooding

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL17

Simulation IssuesDoes not take into account for any delay caused by meta-data negotiationns constraints: Memory CPU timeA simulator model of a real-world system is necessarily a simplification of the real-world system itself

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL18

SPIN-BC MotivationsOne-to-many communication is:1/n times cheaper in a broadcast network than in a point-to-point networkwhere n is the number of neighbours for each nodeSaves energyLets each node overhear all transactions that occur coordinate better

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL19

SPIN-BCFor lossless broadcast networkUses a shared channelLike SPIN-PP, uses ADV, REQ and DATA messagesThree differences: Messages sent to a broadcast address When received ADV, sets random timer, sends

REQ upon timeout. Other nodes hearing REQ will cancel their timer

Nodes will send data to the broadcast address only once, assuming lossless network

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL20

ADVE

DREQ

D

E

SPIN-BC Example

DATA E

DC

ADV

E

DC

BA

A Nodes with dataA Nodes without data

A Nodes waiting to transmit REQ

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL21

SPIN-RLFor lossy broadcast networkTwo modifications Firstly, if a node does not receive data

within a period of time, it sends REQ again

Secondly, when a data item is repeatedly requested, the node will wait for a predetermined amount of time before responding to any requests.

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL22

SPIN - BC and RL : best option?Open questions: Bandwidth-saving, how about utilising IP

Multicast? Reliable multicast?

Need further research Our opinion: if yes, a trimmed-down

version of multicasting is needed.

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL23

Broadcast Media Simulations Simulation Testbed same as the one used in SPIN-PP with following variations:

Single shared-media channel Nodes use 802.11 MAC layer protocol Delay and packet losses taken into account

Simulation Setup monarch – extension of ns

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL24

Simulations with No Packet Losses

--- SPIN-BC--- Ideal--- Flooding

SPIN-BC Converges quicker than flooding Dissipates 50% less energy as compared to

flooding

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL25

Simulations with Packet Losses

--- SPIN-BC--- SPIN-RL--- Ideal-- Flooding-

SPIN-RL Only ideal and SPIN-RL converge because of their ability to

recover from packet loss, rest do not converge This is closer to reality scenario. Expends more energy as compared to BC and the ideal

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL26

Data Distributed Per Unit Energy

SPIN-RL delivers twice as much data per unit energy than flooding (100% more)

--- SPIN-BC--- SPIN-RL--- Ideal--- Flooding

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL27

Validity/Relevance of resultsSimulation environment selected in SPIN-RL is a better representation of real world scenarioChannel interference and collision which were ignored in SPIN-BC, PP and EC have been taken into accountSPIN-RL: Theoretical integrity consistent

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL28

Major Short-comingsSimulation Environment does not closely model Wireless Sensor Networks environmentFalse assumption: the infinite supply of energy in SPIN-RLResults fall short of supporting a convincing argument in favour of SPIN protocols

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL29

Summary of relevant/similar work

What is similar and/or relevant?

SPIN and NNTP – comparable?

SPIN and Energy-Conservation based routing

SPIN and other Flat Multi-hop routing protocols

Spin and Others – AIDA, LEACH

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL30

SPIN vs Directed DiffusionWhat is directed diffusion? Similarities:

Optimized for disseminating application-specific information in a sensor network, specifically between source and sink nodes

Use of data naming allows negotiation between nodes prior to data forwarding to eliminate redundancy

Interest (REQ) and data (DATA) caches maintained at each node

Node-local decision making

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL31

SPIN vs Directed Diffusion - 2Dissimilarities:

SPIN uses a push model for disseminating information to all nodes, while DD uses a pull model for obtaining information

Data is sent to all nodes in SPIN while data is NOT sent to all nodes in Directed diffusion.

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL32

Sensor Network Applications and SPIN

Applications make the Networks SPIN around

Typical Sensor Network Applications

Application/Network type – Time Critical

Application/Network type – Reliable & Re-Usable

What kind of Protocols are optimal ?

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL33

Applications and SPINApplication/Network type – Time Critical

Characteristics Typical example – Seismic Activity Detection SPIN – is it optimal for this type of apps?

Application/Network type – Reliable & Re-usable

Characteristics Typical example – MARS Habitat Monitoring SPIN – is it optimal for this type of apps?

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL34

Summary and Crystal BallThe Potential of Wireless Sensor NetworksThe Future of Wireless Sensor Networks

The Potential of SPINThe Limitations of SPIN The Future of SPIN

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL35

Ask us!We asked Joanna Kulik, one of the SPIN authors..

X-Matrix: “Could you address any SPIN protocol weaknesses (if any?)”

Joanna: “I haven't thought about SPIN in many years.  I'm sure that there are many weaknesses, and that they would be easy to find.  With SPIN we were just trying to lay some initial groundwork in the field.  With anyinitial work, there are hundreds of ways that the workcould be improved.”

X-Matrix TeamMSc Data Communications, Networks and Distributed Systems; Computer Science Department, UCL36

References1. D. Estrin, R. Govindan, J. Heidemann, S. Kumar,

Next century challenges: Scalable coordination in sensor networks, Proc. MOBICOM, 1999, Seattle, 263-270.

2. C. Intanagonwiwat, R. Govindan, , and D. Estrin. Directed diffusion: A scalable and robust communication paradigm for sensor networks. In MobiCOM, Boston, MA, August 2000.

3. Wireless Networks of Devices (WIND) [http://wind.lcs.mit.edu]

4. Praveen Rentala, Ravi Musunnuri, Shashidhar Gandham, Udit Saxena, Survey on Sensor Networks

5. LEACH [http://nms.lcs.mit.edu/projects/leach]