SolarStore: Enhancing Data Reliability in Solar-powered Storage-centric Sensor Networks Yong Yang, Lili Wang, Dong Kun Noh, Hieu Khac Le and Tarek F. Abdelzah.

SolarStore: Enhancing Data Reliability in Solar-powered

Storage-centric Sensor NetworksYong Yang, Lili Wang, Dong Kun Noh, Hieu Khac Le and Tarek F. Abdelzahe

Mobisys 2009

Brian2009/8/17

Outline

• Introduction • Method • Hardware system• Implementation • Performance• Result • Conclusion

Introduction

• WSN in habitat and environment monitoring– Sensors are deployed in remote locales– Limited connectivity– Data need to be stored in the network– Long-term running

• SolarStore– Energy adaptive– Storage reliability mechanism

Motivations

• Energy– How to estimate redundancy energy to enhance

the reliability? • Storage– How to use the redundancy energy to enhance

the reliability?

Implementation

• 9 nodes in the farm of the University of Illinois at Urbana-Champaign (40.1N, 88.20W)

• 12V, 98AH• 120Watts

Hardware

• EEE PC :10~15Watts (0.8~1.2A for 12V), 18GB• Linksys WRT54GL : 2.4Watts– >3Mbps transmission by 50m outdoor

• Phidget voltage sensor:0.06V resolution

Architecture of SolarStore

• Repository: a piece of storage space on the solid state disk managed by the operating system

• Replicator: reads data blocks from Repository and encodes them into data chunks

• Receiver: receives the encoded data chunks from other nodes and stores them into Repository

Architecture of SolarStore

Method

• Eresidual: current residual energy in battery• Tfull(Eresidual):expected time when battery is full• C: battery capacity• Psolar: average power charging rate by solar panel• Psys: average power consumption rate by system

Method

• How to get Eresidual threshold if B(Eresidual)=0? B(Eresidual): the expected duration of blackout time

1. Eresidual = Psys* Tfull(Eresidual) at least2. 3. Eresidual threshold = C*(Psys/Psolar)

• △E: energy allocated for enhancing data reliability(if Eresidual C*(Psys/Psolar) )≧ △E = Eresidual - C*(Psys/Psolar)

Method

• Sresidual: current residual storage space left• △S: storage surplus• R: expected data sensing rate• M: expected time from now to the next

upload opportunity △S= Sresidual - R*M

Data coding and Reliability level

• Fountain coding for replication– partitions a data block into k chunks and generates

k’ (k’ k) encoded chunks, eg. k=8, k’=12≧– Scatter out to each neighbor k’/(g+1) chunks, g= amount of neighbors(eg. g=8)

• Reliability level : α=k’/h • h:the number of data chunks stored on the node that

were generated from this data block

Voltage charging characteristic

• Charging on from 6AM~7PM• 14.0V as 100%• 11.0V as 0%

Performance evaluation

• Charging current from Oct.21~Nov.4 2008

• Emulation

Three Experiments

• Under different energy states• Adaption to other environment• Comparison to three other schemes

Under different energy states

• Residual energy• the behavior of SolarStore in a long run does

not depend on the initial states

Under different energy states

• Residual storage and storage surplus• Surplus remain constant

Node 2 Node 9

Adaption to other environment

• Enlarge charging current by 3 times for one day every 3 three days

• The other two days multiply 0.2

Comparison to three other schemes

• 0-Reliable – no data replication at all and uses all energy and

storage space for data sensing• 1-Reliable – always replicates data to maximize data reliability

• full-Reliable – only starts data replication when the battery is

nearly full (99%) because the energy charged from solar panels will be wasted if not used.

Comparison to three other schemes

• Data loss– Data sensing during energy blackout– Node failure

• 0-Reliable :worst at node failure• 1-Reliable : best at recovering• full-Reliable : at least 58% data loss

Conclusion

• the behavior of SolarStore in a long run does not depend on the initial states

• SolarStore can dynamically responds to variations in the environment

• leads to more retrievable data under different node failure scenarios, compared to three other schemes

• Pros– Adaptive to control energy and storage effectively

• Cons– Not consider the severe weather deeply– How to coordinate energy sharing between

Replicator and Receiver?

• Thank you

Reliability level of node 9