wireless sensor network performance monitoring · 2007-07-13 · data. challenges limited energy...
TRANSCRIPT
Wireless Sensor Network Wireless Sensor Network Performance MonitoringPerformance Monitoring
Yaqoob J. AlYaqoob J. Al--raisiraisi & David J. Parish& David J. ParishHigh Speed Networks GroupHigh Speed Networks Group
Loughborough UniversityLoughborough University
MSN MSN CosenersCoseners
1212--13th July 200713th July 2007
OverviewOverview
�� The problem we are trying to solve and why it The problem we are trying to solve and why it is exist.is exist.
�� How it has been solved by others and How it has been solved by others and solutionssolutions’’ limitations.limitations.
�� The proposed algorithm characteristics, The proposed algorithm characteristics, novelty and function.novelty and function.
�� Empirical and simulation experimental Empirical and simulation experimental results.results.
�� Conclusion and future work.Conclusion and future work.
Wireless Sensor Networks (Wireless Sensor Networks (WSNsWSNs))
Characteristics Characteristics
�� Small, disposal, low power consumption Small, disposal, low power consumption network nodes.network nodes.
�� Reduce network and deployment cost.Reduce network and deployment cost.�� Enable network to be deployed in any Enable network to be deployed in any
terrain.terrain.�� Extend the virtual functionality of Extend the virtual functionality of
traditional networks.traditional networks.�� Increase the reliability of network collected Increase the reliability of network collected
data.data.
ChallengesChallenges
�� Limited energy and communication Limited energy and communication resources.resources.
�� No dominant protocols for all network No dominant protocols for all network applications.applications.
�� Frequent changes in connectivity, link Frequent changes in connectivity, link failure and the change node status.failure and the change node status.
�� Unbalanced network traffic.Unbalanced network traffic.�� Change the degree of networkChange the degree of network’’s tolerance s tolerance
to data changes and losses.to data changes and losses.
Deviation of nodesDeviation of nodes’’ operation operation impactimpact
�� Use more network resources and reduce Use more network resources and reduce network lifetime.network lifetime.
�� Reduce the reliability of collected data.Reduce the reliability of collected data.
Related workRelated work
�� FaultFault--Tolerant ToolsTolerant Tools�� Data Cleaning ToolsData Cleaning Tools�� Diagnosis toolsDiagnosis tools�� Performance Performance
monitoring/measuring monitoring/measuring ToolsTools
�� Used only one level Used only one level parameterparameter
�� Definition of outlier/faultDefinition of outlier/fault�� Used method (either Used method (either
complex or not complex or not considering loss)considering loss)
�� Consume resources in Consume resources in packet exchangepacket exchange
�� Clear data without Clear data without indicating the confidence indicating the confidence levellevel
�� Map of common Map of common neighbourhood functionneighbourhood function
Main method limitationsMain method limitations
�� The impact on network lifetimeThe impact on network lifetime
�� Tracking of crash faults.Tracking of crash faults.
�� Tracking of either applicationTracking of either application-- or or communicationcommunication--level parameters.level parameters.
�� The generated warning packets.The generated warning packets.
Main study objectiveMain study objective
�� To develop a low resource usage tool that To develop a low resource usage tool that tracks network nodestracks network nodes’’ operation deviation operation deviation ‘‘onon--lineline’’ before they degrade collected before they degrade collected data reliability and network lifetime.data reliability and network lifetime.
AlgorithmAlgorithm
Algorithm Characteristics and Algorithm Characteristics and noveltynovelty
�� The utilization of all neighbour nodesThe utilization of all neighbour nodes’’ information in information in range.range.
�� The extraction of its metrics from network protocolsThe extraction of its metrics from network protocols’’stored parameters.stored parameters.
�� The use of a simple statistical method with tolerance The use of a simple statistical method with tolerance to loss.to loss.
�� The definition of tracked deviation.The definition of tracked deviation.�� The relation between high and low network levels by The relation between high and low network levels by
testing the distortion.testing the distortion.�� The test of possible communication available.The test of possible communication available.�� The positive tracking of neighbour detection and The positive tracking of neighbour detection and
control of warning packet release.control of warning packet release.
Algorithm event detectionAlgorithm event detection
�� Node malfunction.Node malfunction.
�� Neighbourhood malfunction.Neighbourhood malfunction.
�� Coverage problem.Coverage problem.
�� Inefficient power consumption.Inefficient power consumption.
�� Communication problem.Communication problem.
Algorithm performance evaluationAlgorithm performance evaluation
�� Lightweight operation (energy Lightweight operation (energy consumption, algorithm analysis)consumption, algorithm analysis)
�� Robustness (tested under different packet Robustness (tested under different packet loss, deviated nodes, deviated loss, deviated nodes, deviated measurements, dead nodes)measurements, dead nodes)
�� Adaptability and responsiveness (empirical Adaptability and responsiveness (empirical experiments in single and multiexperiments in single and multi--hop hop configuration).configuration).
�� Scalability.Scalability.
Algorithm DetectionAlgorithm Detection
0 10 20 30 40 50 60 70 80 900
10
20
30
40
50
60
70
80
90
100
Percentage of faulty nodes
Det
ectio
n pe
rcen
tage
Average positive false detection Average deviated nodes detected Average non detected deviated nodes
020406080
100120
10 20 30 40 50 60
Neighbours Fault Percentage
Per
cen
tag
e 4%8%13%16%22%25%30%
0 1000 2000 3000 4000 5000 6000 7000 80000
1
2
3
4
5
6
7
8
9
10
Event number
Per
cent
age
of c
hang
e
Node1Node 2Node 3Node 4Node 5Node 6Node 7Node 8Node 9Node10Node 11
Temporal deviation
Coverage problem
Permanent deviation
Threshold value
Conclusion and future workConclusion and future work
�� The experiments showed the efficiency of the The experiments showed the efficiency of the passive proposed algorithm in detecting the passive proposed algorithm in detecting the deviations and low power consumption.deviations and low power consumption.
�� The experiments detect degradation of the The experiments detect degradation of the algorithm functionality in very small networks.algorithm functionality in very small networks.
�� Solve this limitation by cause and effect Solve this limitation by cause and effect diagram.diagram.
�� Test the algorithm in different network Test the algorithm in different network applications other than environment monitoring.applications other than environment monitoring.
�� Test the effect of node mobility on the proposed Test the effect of node mobility on the proposed algorithm performance.algorithm performance.
Many ThanksMany Thanks
Why it existsWhy it exists
�� Low manufacturing material quality and Low manufacturing material quality and process.process.
�� Limited usage of faultLimited usage of fault--tolerant and diagnosis tolerant and diagnosis techniques.techniques.
�� Harsh environment the nodes operate in.Harsh environment the nodes operate in.
�� Direct coupling with monitored phenomenon.Direct coupling with monitored phenomenon.
�� Usage of wireless medium.Usage of wireless medium.
Type of deviationType of deviation
�� Systematic: affect the operation cautiously Systematic: affect the operation cautiously until the problem is solved such as until the problem is solved such as calibration, reduction in operating power.calibration, reduction in operating power.
�� Transient: affect the operation until the Transient: affect the operation until the effect disappears such as random effect disappears such as random environmental effects and unstable environmental effects and unstable characteristics of hardware.characteristics of hardware.
Effect on the quality and quantityEffect on the quality and quantity
��Usage of more resources. Usage of more resources.
��Reduce the quality or quantity Reduce the quality or quantity of collected data. of collected data.
��Sensor node measurement Sensor node measurement such as biased, drift and stuck.such as biased, drift and stuck.
��Dropping exchange packets.Dropping exchange packets.
��Reboot or stop node function.Reboot or stop node function.
Indirect effectIndirect effectDirect effectDirect effect
The problem we are trying to solveThe problem we are trying to solve
Maintain the large scale Wireless Sensor Maintain the large scale Wireless Sensor NetworkNetwork’’s collected data quality and s collected data quality and quantity at acceptable level.quantity at acceptable level.
Listening and filtering moduleListening and filtering module
�� Integrate the algorithm Integrate the algorithm functionality into network functionality into network application flow process.application flow process.
�� Use network applicationUse network application’’s s existing parameters.existing parameters.
�� Filter the high deviated Filter the high deviated parameters.parameters.
Data analysis and threshold test Data analysis and threshold test modulemodule
�� Test median value validityTest median value validity�� Calculate the residual of Calculate the residual of
each neighbour in each neighbour in neighbourhood.neighbourhood.
�� Test each residual value Test each residual value to predefined threshold.to predefined threshold.
�� Test the validity of data.Test the validity of data.
Decision confidence control moduleDecision confidence control module
�� Calculate the Calculate the frequency of frequency of deviation existence deviation existence with the tolerance with the tolerance of the network of the network application protocolapplication protocol
�� Request from Request from module 4 to send module 4 to send warning message.warning message.
Warning packet exchange moduleWarning packet exchange module
�� Test if any neighbour Test if any neighbour sends the same sends the same message.message.
�� Test received neighbour Test received neighbour warning message.warning message.
�� Send warning detection Send warning detection messagemessage
�� Control packet releasingControl packet releasing�� Request the protocols to Request the protocols to
reconfigure the node reconfigure the node setting.setting.
0 2000 4000 6000 80000
20
40
60
80
100Figure 1: Performance
Event number
Per
cent
age
of c
hang
e
Net. pref.Reding pref.
0 2000 4000 6000 8000-1
-0.5
0
0.5
1Figure 2: Percentage of networtk losses
Event number
Per
cent
age
of c
hang
e
Per. of loss
0 2000 4000 6000 800010
15
20
25
30Figure 3: Median
Event number
Deg
rees
cen
tigra
de
median
0 2000 4000 6000 80000
20
40
60
80
100Figure 4: Residual changes
Event number
Per
cent
age
of c
hang
e
0 2000 4000 6000 80000
20
40
60
80
100Figure 1: Performance
Event number
Per
cent
age
of c
hang
e
Net. pref.Reding pref.
0 2000 4000 6000 8000-1
-0.5
0
0.5
1Figure 2: Percentage of networtk losses
Event number
Per
cent
age
of c
hang
e
Per. of loss
0 2000 4000 6000 800010
15
20
25
30Figure 3: Median
Event number
Deg
rees
cen
tigra
de
median
0 2000 4000 6000 80000
10
20
30
40Figure 4: Residual changes
Event number
Per
cent
age
of c
hang
e
0 2000 4000 6000 80000
20
40
60
80
100Figure 1: Performance
Event number
Per
cent
age
of c
hang
e
Net. pref.Reding pref.
0 2000 4000 6000 80000
10
20
30
40Figure 2: Percentage of networtk losses
Event number
Per
cent
age
of c
hang
e
Per. of loss
0 2000 4000 6000 800010
15
20
25
30Figure 3: Median
Event number
Deg
rees
cen
tigra
de
median
0 2000 4000 6000 80000
10
20
30
40
50Figure 4: Residual changes
Event number
Per
cent
age
of c
hang
e
0 2 4 6 8 10
x 104
0
20
40
60
80
100Figure 1: Performance
Event number
Per
cent
age
of c
hang
e
Net. pref.Reding pref.
0 2 4 6 8 10
x 104
0
20
40
60
80
100Figure 2: Percentage of networtk losses
Event number
Per
cent
age
of c
hang
e
Per. of loss
0 2 4 6 8 10
x 104
0
50
100
150Figure 3: Median
Event number
Deg
rees
cen
tigra
de
median
0 2 4 6 8 10
x 104
0
2
4
6
8
10Figure 4: Residual changes
Event number
Per
cent
age
of c
hang
e
Relating the high and low Relating the high and low parametersparameters
6 6.1 6.2 6.3 6.4 6.5 6.6 6.7
x 104
0
20
40
60
80
100
120
140
Event number
Val
ue in
deg
rees
cen
tigra
de
Method 1Method 1
Method 2Method 2
Algorithm warning packetAlgorithm warning packet