1. REED : Robust, Efficient Filtering and Event Detection
in Sensor Networks
Daniel J. Abadi, Samuel Madden, and Wolfgang Lindner
MIT CSAIL
31st VLDB Conference, 2005

2. Contents What problem? Constraints : Sensor Networks
Sensor Database Motivation
Naïve Join Algorithm
Ideal Join Algorithm
REED Algorithm
Bloom Filter Optimization
Conclusion

3. What Problem?
Complex data filtering in sensor networks. 1

4. Constraints : Sensor Networks
Sensor nodes are small, battery-powered devices
7Mhz processor
38.6Kbps radio with ~100 feet range
4K RAM, 128K Program Flash, 512K Data Flash
Power conservation is important
sensing and transmitting data typically dominate power usage
Berkeley Mote 2

5. Sensor Database Motivation
Programming application is hard
Limited power budget
Lossy, low bandwidth communication
Required long-lived, zero admin deployments
Distributed Algorithms
Limited tools, debugging interfaces
Solution : database style interface(e.g. Tiny DB) 3

6. Naïve Join Algorithm
Send all tuples from data table to root : perform join at root A X B B
Root 0 C D
Predicate Table 1 2 X 3 4 5 X 6 7 4

7. Ideal join Algorithm X X Send join table to each node
At node, perform join
Problem : severe node
memory constraints
Root 0 A A B B C 1 2 C D D A B X B X X C A A 3 4 5 D B B C C D D A A 6 7 X B B X C C D D 5

8. REED Algorithm Cluster nodes into groups Store portion into predicateB A
Cluster nodes into groups
Store portion into predicate
table in each group member
Send sensor data tuples to
every member of group B C B C C D D D
Root 0 1 2 X X 3 4 5 X X X X D X 8 6 7 6

9. Group Formation (1/2) Root 0 7 1 2 3 4 5 6 7 {1,2,5} {1,2,3,4}
{1, 3, 4, 6} 3 4 5
{5,2,6,7}
{1,3,4}
{7,5,6} 6 7
{6,5,7, 4} 7

10. Group Formation (2/2) 1 2 3 4 Space: 10 CurrList: {1, 3, 4}
Potential: {1, 3, 4}
Space: 8
CurrList: {1, 4}
Potential: {1, 3, 4}
Space: 4
CurrList: {1}
Potential: {1, 2, 3, 4} 1
Neighbor list: {1, 2, 3, 4} 2
Group Accepted:
{1, 3, 4}
Choose Me!
{1, 3, 4, 6}
Space: 4
Broadcast: Want to make group
Choose Me!
{1, 3, 4}
Space: 2
Neighbor list:
{1, 2, 5} 3 4
Neighbor list: {1, 3, 4, 6}
Neighbor list: {1, 3, 4} 8

11. Bloom Filter Optimization
Might produce false positives but never false negatives
Can be used in conjunction with previous REED algorithm
Step 1: Hash domain of sensor
values onto Bloom Filter
Step 2: Send Bloom Filter to Each Sensor Node
Root 0
Temp: 20 1
hash 2
Temp: 90 1
hash 1 5
Bloom Filter 3 4 6 7 X X 9

12. Experiments and Results
Ran experiments both in simulation and on real motes
For simulation, 40 sensor nodes arranged in a grid
Use TinyOS Packet Level Simulation
Models CSMA backoff
Carrier sense packet delivery model
- Overlap between 2 receptions leads to both being corrupted
Use TinyOS MintRoute for MultiHop Routing Layer 10

13. REED performs Well at most Selectivities11

14. Simulated Results Match Real Results From motes
Ran REED algorithm on a simple 5 node sensor network 12

15. Conclusion (1/2) Contributions
Complex filters → table of expressions → join
REED algorithms capable of
- Running with limited amounts of RAM
- Robustness in the face of message loss and node failure
Experiments show benefits of doing complex join-based filters
in the sensor network 13

16. Conclusion (2/2) Motivating Applications Industrial Process control
- Distributed sensors measure environmental variables
- Want to know if exceptional condition is reached
Failure and Outlier Detection
- Look for de-correlated sensor readings
Power scheduling
- Minimize power consumption by distributing work across sensors 14