distributed monitoring and aggregation in wireless sensor networks infocom 2010 changlei liu and...
Post on 03-Jan-2016
219 Views
Preview:
TRANSCRIPT
Distributed Monitoring and Aggregation in Wireless Sensor Networks
INFOCOM 2010
Changlei Liu and Guohong CaoSpeaker: Wun-Cheng Li
Outline• Introduction• Goal• Distributed Poller Selection Algorithms
▫Randomized Algorithm▫Deterministic Algorithm▫Hybrid Algorithm
• Performance evaluation• Conclusion
2
Introduction
•As sensor nodes usually operate in an unattended harsh environment, they are prone to failure and may run out of battery
•To make sensor network reliable as well as adaptable, sensor status has to be closely monitored▫liveness▫density estimation▫residue energy
3
Introduction
• In distributed systems, the only way to learn the status of a node is through receiving messages from the node▫Poller-Pollee structure has been widely used for
network management
4
Introduction
•Compared with the wired networks, designing monitoring mechanisms for sensor networks has more challenges.▫Dynamic topology▫sensors need to self-organize themselves into a
monitoring architecture
5
• If the number of pollers is too small then false alarm rate may increase as a consequence▫pollees will be too far away from the poller
Problem
6
•To reduce the monitoring overhead, we take the hop-by-hop aggregation opportunities in sensor networks.
Problem
7
poller poller poller
s : aggregation ratio
Goal
•Strike a balance between the number of Pollers and false alarm rate▫Minimum Poller Selection
8
•Two widely used operational modes of the poller-pollee structure
System model
9
Poller
Pollee
2reply/s 2poll/s
Poller
Pollee
2reply/s
System model
• The poller-pollee based monitoring.▫Failure rate fi
10
Td : detection time
t: polling time interval
Distributed Poller Selection Algorithms
•Randomized Algorithm▫Each node elects itself as a poller with probability ρ
11
1
2
9
10
3
7
5
4
6
11
8
2
9
10
3
7
5
4
11
8
6
1
poller
pollee
Unlabeled node
Distributed Poller Selection Algorithms
•Deterministic Algorithm▫Uses two parameters k1, k2 to guide a better distribution
of poller and pollee▫No two pollers are less than k1hops away from each
other▫No pollee is more than k2 hops away from its poller.
12
Distributed Poller Selection Algorithms
•Deterministic Algorithm▫ k1=k2=1
13
1
2
9
10
3
7
5
4
6
11
8
poller
pollee
Unlabeled node
2
9
10
7
5
4
6
11
8
1
3
2
9
10
7
4
6
8
3
5 11
1
Distributed Poller Selection Algorithms
•Hybrid Algorithm▫ k1=k2=1
14
1
2
9
10
3
7
5
4
6
11
8
poller
pollee
Unlabeled node
3
1
9
10
7
5
4
11
8
2 6
1
9
10
7
5
4
6
11
8
2
3
1
9
10
7
4
6
8
2
3
5 11
Performance evaluation
•C++
15
Parameter SettingsRandomly Deployed region 20 × 20Nodes 1000~1500transmission range 1Failure rate fi 0.05Detection time Td 2t
Performance evaluation
16
Performance evaluation
17
Performance evaluation
•=0.2•K1=1,k2=2
18
Randomized algorthm Hybrid algorthm
Performance evaluation
•=0.75, 0.065•K=6
19
Randomized algorthm Hybrid algorthm
Performance evaluation
20
Conclusions
•This paper proposed a fully distributed algorithms to select the minimum number of pollers while bounding the false alarm rate.
•Simulations results the hybrid algorithm can reduce the message overhead significantly
21
Thank you!
22
top related