The document proposes STDCS, a spatio-temporal data-centric storage scheme for real-time sensor network applications. STDCS uses a novel spatio-temporal indexing technique to dynamically balance the storage load across sensors based on query hotspots. It assigns local virtual addresses, delivers readings and processes queries using the spatial index. STDCS adapts to changing hotspots over time through techniques like hotspot decomposition and switching time adjustment, improving metrics like throughput, energy usage and node deaths compared to alternatives like local storage or spatial indexing alone.

1. STDCS: A Spatio-Temporal Data-Centric Storage Scheme For Real-Time Sensornet Applications Mohamed Aly (University of Pittsburgh & Yahoo, Inc.) In collaboration with Anandha Gopalan (University of Pittsburgh, Imperial College) and Jerry Zhao, Adel Youssef (Google, Inc.)

2. Motivation: Real-Time Geo-Centric Sensor Network Applications Globally deployed sensor around the globe. Clusters of sensors forming networks. Mobile users roaming across the networks. Real-time geo-centric ad-hoc queries issued from within or nearby the queried area. The sensor network is responsible of answering these queries directly from the sensors rather than from base stations. Examples: Bronx Zoo cluster. Disaster management cluster.

3. Motivation: Real-Time Geo-Centric Sensor Network Applications

4. Data Storage Options in Sensor Networks Base Station Storage: Events are sent to base stations where queries are issued and evaluated. Best suited for continuous queries. In-Network Storage (INS): Events are stored in the sensor nodes. Best suited for ad-hoc queries. All previous INS schemes were Data-Centric Storage (DCS) schemes.

5. In-Network Data-Centric Storage (DCS) Mainly to answer range queries. Quality of Data (QoD) of ad-hoc queries. Assign a value-range of readings for each sensor. Examples: Distributed Hash Tables (DHT) [Shenker et. al., HotNets’02] Geographic Hash Tables (GHT) [Ratnasamy et. al., WSNA’02] Distributed Index for Multi-dimensional data (DIM) [Li et. al., SenSys’03, Aly et. al., DMSN’05, MOBIQUITOUS’06] K-D Tree based Data-Centric Storage (KDDCS) [Aly et. al., CIKM’06]

6. STDCS Overview Motivation: No previous INS schemes adopting geo-centric storage. Expected techniques may be: Local storage. Spatial storage Design Goal: Load-Balancing of storage load among sensors Differences from previous schemes: Temporally evolving spatial indexing scheme to balance query load among sensors. Dynamic query hotspot detection and decomposition.

7. Roadmap Motivation: Real-Time Geo-Centric applications. Background: Data-Centric Storage (DCS). Problem Statement: Real-Time Geo-Centric Storage. Scheme Overview: STDCS. STDCS Components Local Virtual address assignment Spatio-Temporal data indexing. Point-to-point data delivery. Query processing. Adaptive hotspot decomposition. Experimental Results Conclusions

8. STDCS Components: Local Virtual Address Assignment

9. STDCS Components: Spatio-Temporal Data Indexing

10. STDCS Components: Reading Delivery and Querying

11. STDCS Components: Adaptive Hotspot Decomposition Motivation: Dynamic query hotspots as time progresses. Observation: Recurrent querying scenarios across the day, the week, etc. Technique: Continuously keeping track of hotspots using the Average Querying Frequency (AQF) metric. Dynamically chaning the switching time to decompose hotspots.

12. Roadmap Motivation: Real-Time Geo-Centric applications. Background: Data-Centric Storage (DCS). Problem Statement: Real-Time Geo-Centric Storage. Scheme Overview: STDCS. STDCS Components Local Virtual address assignment Spatio-Temporal data indexing. Point-to-point data delivery. Query processing. Adaptive hotspot decomposition. Experimental Results Conclusions

13. Simulation Description Compare: STDCS, local storage, spatial indexing. A cluster of stationary sensors (with random locations). Each sensor senses a reading each 10 min. Sensor reading = 1 packet. Sensor capacity = 20 readings (packets) Multiple mobile users. A query: random sensor, radius, and type. Two phases: initialization (3 hours of readings) & running (1 day of readings and queries). Metrics: throughput, energy level, node deaths.

14. Experimental Results: STDCS vs. Query Hotspots

15. Experimental Results: STDCS vs. Query Hotspots

16. Experimental Results: Switching Time Effect

17. Experimental Results: Switching Time vs. Node Deaths

18. Experimental Results: Adaptive Hotspot Decomposition

19. Conclusions STDCS: A real-time geo-centric data storage scheme. A new concept of spatio-temporal data indexing. Ability to dynamically cope with dynamic loads and query hotspots.

20. Acknowledgment This work has been partly supported by: Google, Inc. The “Secure CITI: A Secure Critical Information Technology Infrastructure for Disaster Management (S-CITI)” project funded through the ITR Medium Award ANI-0325353 from the National Science Foundation (NSF). For more information, please visit: http://www.cs.pitt.edu/s-citi/

21. Thank You Questions ?