300 likes | 472 Views
PRESTO: Feedback-driven Data Management in Sensor Network. (* PRE dictive STO rage ). Ming Li, Deepak Ganesan, and Prashant Shenoy University of Massachusetts, Amherst. Tracking. Structure/Machinery Monitoring. Emerging large-scale sensor networks.
E N D
PRESTO: Feedback-driven Data Management in Sensor Network (*PREdictive STOrage) Ming Li, Deepak Ganesan, and Prashant Shenoy University of Massachusetts, Amherst
Tracking Structure/Machinery Monitoring Emerging large-scale sensor networks • Hierarchical wireless networks composed of low power sensors. • Enables densely and closely monitoring of phenomena. Surveillance
Hierarchical Sensor Network Architecture Remote Sensors Remote Sensors Sensor Proxy Sensor Proxy Mesh Network Client Data Browsing, Querying and Processing Base-station Internet
Sensor-centric Architecture Proxy-centric Architecture Approaches to Proxy-Sensor Interaction
Proxy-Centric Architecture • Overview • Proxy determines when to pull data, which sensor to query, and what data to pull using complex modeling and query processing mechanisms. • Pros: • Intelligence placed where resources are available. • More complex algorithms possible. • Cons: • Cannot capture anomalies. • Less energy-efficiency • Greater query error. BBQ [Deshpande04]
Sensor-Centric Architecture • Overview • Forward queries into the sensor network. Perform data fusion, query processing and filtering within the network. • Pros: • Greater query accuracy • Better energy-efficiency. • Cons: • Greater sensor complexity. • Greater query latency. Directed Diffusion [Heidemann01]
Sensor-centric Proxy-centric PRESTO Model PRESTO
Key Ideas in PRESTO • Steal from the rich (proxy) and give to the poor (sensors). • Exploit predictable structure in sensor data when possible. • Adapt to data & query dynamics to minimize energy usage. • Exploit low-power storage for efficient archival querying.
Outline • Motivation • Key Ideas • Example • ARIMA Model • Evaluation • Summary & Future Work
Model Build Model Data Example-Modeling Sensor Proxy
Yes Predict Predict Example-Model Driven Push Proxy Sensor
Query What is the reading at time t with confidence c? No Pull Tt Yes Example-Query Proxy Sensor
Model Build Model Example-Feedback Sensor Proxy
Interpolation Interpolation Push Tt Example - Update Cache after Push Sensor Proxy
Re-prediction Re-prediction Pull Tt Interpolation Interpolation Example - Update Cache after Pull Proxy Sensor
Outline • Motivation • Key Ideas • Example • ARIMA Model • Evaluation • Summary & Future Work
Goals • Catches data trends • Easy to compute on sensors
Data Trends • Temperature data trace shows very obvious temporal trend • Shows both long term trend and short term trend. Seasonal Period
Data Trends • ARIMA model can catch both of these trends Long Term Trend Short Term Trend
Computation • Easy to predict • Five additions and three multiplies Previous prediction results Previous prediction errors
Outline • Motivation • Key Ideas • Example • ARIMA Model • Evaluation • Summary & Future Work
Evaluations • Both numerical simulations and real deployments • Test Bed: • 1 Stargate (Proxy) / 20 Tmote’s (Sensor) • 1 Stargate acts as emulator • Data Trace: • James Reserve
Micro Benchmark Model Asymmetry Cost of model building is 500x more than prediction Breakdown of Energy Costs Total cost of prediction and storage is 10x less than communication.
Model-driven Push Performance • Matlab simulation shows that Model-driven push performs better than model-driven pull.
Scalability • Impact of System Scale • Uses emulator to get large network scale Support up to 100 sensor nodes per proxy
Scalability • Impact of Query Frequency • System adapts to high query frequency. • Query latency does increase with query frequency Most of the queries are answered using proxy cache
Adaptation • Adapt to query dynamics • Reduce query latency by 50% compared to before adaptation Average query tolerance changes to a lower value which brings more pulls Adapt to the low query tolerance after a short period
Adaptation • Adapt to data dynamics • Reduce communication by 30% compared to non-adaptive scheme Reduces 30% of communications
Failure Detection • Detect sensor failure using pulling messages • Detection latency decreases with query interval, as well as query tolerance. Longest detection latency less than 2 hours