[email protected] & [email protected] elham hormozi & razieh asadi university of...

31
[email protected] & [email protected] FAULT TOLERANCE IN WIRELESS SENSOR NETWORK Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

Upload: barry-fletcher

Post on 25-Dec-2015

217 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

1

[email protected] & [email protected]

FAULT TOLERANCE

IN WIRELESS SENSOR

NETWORK

Elham Hormozi & Razieh AsadiUniversity of Science & Technology Mazandaran Babol

Page 2: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

2

Outline

Review of Wireless Sensor Network

Fault Tolerance in WSNs

Fault Detection

Fault Recovery

Relay Node Placement in Wireless Sensor

Networks

Hop-by-Hop TCP for Sensor Networks

Conclusion

Page 3: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

3

Review of Wireless Sensor Network

A WSN is a self-organized network that consists of a large number of low-cost and low powered sensor devices, called sensor nodes

Can be deployed on the ground, in the air, in vehicles, on bodies, under water, and inside buildings

Each sensor node is equipped with a sensing unit, which is used to capture events of interest, and a wireless transceiver, which is used to transform the captured events back to the base station, called sink node

Sensor nodes collaborate with each other to perform tasks of data sensing, data communication, and data processing

Page 4: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

4

Type of failure in WSNs

Energy depletion Have very limited energy and their batteries cannot usually be recharged

or replaced, due to hostile or hazardous environments Hardware failure

A sensor node has two component: sensing unit and wireless transceiver Usually directly interact with the environment, which is subject to variety of

physical, chemical, and biological factors. Communication link errors

Even if condition of the hardware is good, the communication between sensor nodes is affected by many factors, such as signal strength, antenna angle, obstacles, weather conditions

Malicious attack

It results in low reliability of performance of sensor nodes.Therefore, fault tolerance is one of the critical issues in WSNs

Page 5: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

5

Fault Detection:

Centralized Approach• Sympathy• Secure Locations

Distributed Approach1. Node Self-detection2. Clustering Approach( MANNA)

Page 6: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

6

Sympathy[4]

Using a message-flooding approach to pool event data and current states (metrics) from sensor node

Nodes periodically send metrics back to a sink to detect failures and cause of failure

Given sensor hardware and network limitations, these transmitted metrics must be minimized

Insufficient data at the sink implies failure; sufficient data at the sink implies acceptable network behavior

Based on these metrics, it detects which nodes or components have not delivered sufficient data and infers the causes of failures

Page 7: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

7

Secure Locations[5]

Work on location-aware sensor networks

Introduces a scalable trust-based routing protocol (TRANS)

Select trusted paths that do not include misbehaving nodes by identifying the insecure locations and routing

Include two parts:1. trust routing 2. insecure location discovery and isolation

Page 8: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

8

Secure Locations (cont’d)

Select a secure path and avoid insecure locations

All destination nodes use TESLA, to authenticate all requests

1. sink creates a message with( source location, destination location, authentication message)

2. encrypts this message with its share key and broadcasts it.3. neighbors who know its shared key will be able to decrypt

the request

4. trusted neighbor decrypts the request, adds its location, encrypts the message with its share key and sends it to neighbors

Page 9: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

9

Secure Locations (cont’d)

Use Expanding TTL Search (ETS).

1. Sink marks data packets with increasing hop-count

2. Each intermediate node decrements the hop-count before forwarding

3. When hop count reaches zero node sends ACK to the source informing it of its location is safe

4. The source identifies that part of the path as safe and increases the hop count in subsequent packets.

Page 10: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

10

Advantage & Disadvantage of Centralize Approaches

The centralized approach is efficient and accurate to identify the network faults in certain ways

Resource-constrained sensor networks can not always afford to periodically collect all the sensor measurements and states in a centralized manner

Central node easily becomes a single point of data traffic concentration in the network, as it is responsible for all the fault detection and fault management

This subsequently causes a high volume of message traffic and quick energy depletion in certain regions of the network, especially the nodes closer to the base station

Page 11: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

11

Advantage & Disadvantage of Centralize Approaches(cont’d)

This approach will become extremely inefficient and expensive in consideration of a large-scale sensor network

Multi-hops communication of this approach will also increase the response delay from the base station to faults occurred in the network

Therefore, we have to seek a localized and more Computationally efficient fault detection model

Page 12: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

12

Distributed Approach & Node Self-detection

Use flexible circuit acts as a sensing layer around a node, capable of sensing the physical condition of a node.

Detect physical faults requires the use:

1. Hardware interface consists of a sensing layer(wraps around the node).

2. Software interface reads the sensors, and transmits the data to the Sink

Use TinyOS( have very small footprint, energy-aware, event-based )

Figure 1: SYS25 node.

Page 13: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

13

Distributed Approach & Clustering Approach MANNA Design for event-driven WSN

Clustering use for building scalable and energy balanced applications for WSNs

Distribute fault management into each cluster

Management agents execute in the cluster-heads

This mechanism decreases the information flow and energy consumption as well

A manager is located externally to the WSN has a global vision

Page 14: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

14

Distributed Approach & Clustering Approach MANNA Management application is divided into two phases:

Installation

Occurs as soon as the nodes are deployed in the

network.

Each node report its position and energy to the agent

located in the cluster-head.

Agent sends a LOCATION TRAP and ENERGY TRAP to

the manager

Manager build topology map model and the WSN

energy model

Page 15: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

15

Distributed Approach & Clustering Approach MANNA Management application is divided into two phases:

Operation

Each node report its energy level and position to the

agent whenever there is a state change (another

ENERGY TRAP or LOCATION TRAP)

Manager rebuild topology map model and energy

model

Manager sends GET operations in order to retrieve

the node state

Page 16: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

16

Fault Recovery

WSN restructured or reconfigured, in such a way that failures or faulty nodes do not impact further on network performance

The most commonly used technique for fault recovery is replication or redundancy of components that are prone to be failure When some nodes fail to provide data, the base

station still gets sufficient data if redundant sensor nodes are deployed in the region

Page 17: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

17

Fault Recovery(cont’d)

Relay Node Placement in Wireless Sensor Networks Two-Tiered Wireless Sensor Networks

Hop-by-Hop TCP for Sensor Networks

RideSharing: Fault Tolerant Aggregation

Page 18: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

18

Relay Node Placement in Wireless Sensor Networks(Two-Tiered Wireless Sensor Networks)

Improving reliability and prolonging lifetime of WSNs

Energy consumption is proportional to d for transmitting over distance d, where is a constant in the interval , long distance transmission in WSNs is costly

Employs some powerful relay nodes whose main function is to gather information from raw data from sensor nodes and relay the information to the sink

Relay nodes serve as a backbone of the network

The relay nodes are more powerful than sensor nodes ( energy storage, computing, and communication capabilities)

Page 19: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

19

Two-Tiered Wireless Sensor Networks Each cluster has only one cluster head and each

sensor belongs to at least (backup cluster heads)

Receiver of a relay node fails Data sent by the sensors will be lost Sensor to be reallocated to other cluster heads

Handle general communication faults There should be at least two node-disjoint paths

between each pair of relay nodes in the network

Page 20: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

20

Two-Tiered Wireless Sensor Networks

An intuitive objective of relay node placement in

two-tiered WSNs is to place the minimum number

of relay nodes, such that some degree of fault

tolerance can be achieved.

There are other works that study placement of sensor nodes to make a sensor network k-connected

Page 21: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

21

Hop-by-Hop TCP for Sensor Networks

Why conventional TCP protocol can not be used? Communication links in a sensor network are

unstable TCP protocol over a high loss rate will suffer from

severe performance degradation Sensor may not have sufficient computing power to

implement the entire TCP/IP protocol Hop-by-Hop TCP for Sensor Networks

Aiming to accelerate reliable packet delivery Minimizing end-to-end packet delivery time without

too much throughput degradation Minimizing the number of retransmissions

Page 22: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

22

Hop-by-Hop TCP for Sensor Networks

Every intermediate node execute a light-weight local TCP

Include two part:1. End-to-End TCP

Working on the source and destination nodes

2. One-Hop TCP Working on every node The sender module of a One-Hop TCP is working at

the sender end of a link, and the receiver module is working at the receiver end.

Page 23: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

23

Hop-by-Hop TCP for Sensor Networks

Figure2. Protocol Stack Hop by Hop TCP

Page 24: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

24

End-to-End TCP

Reuse an existing popular TCP protocol, NewReno, with several modifications

1. Sender module forwards packets to the One-Hop TCP module

2. Receiver module receives packets from the One-Hop TCP module

3. One-Hop TCP in each node forwards data packets hop by hop

4. End-to-End ACKs, are forwarded to the source node using One-Hop TCP in the opposite direction

5. Set a larger initial RTO value

Page 25: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

25

One-Hop TCP

A light-weight version of TCP running on each node to forward received packets reliably

Many TCP features, such as packetization and congestion control, are removed

1. Add the IP address of current node to the packet header (receiver knows where to send Local ACK)

2. Set CWND to 13. Set the upper threshold for the number of

retransmissions.

Page 26: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

26

RideSharing: Fault Tolerant Aggregation

Aggregation use for filter redundancy and reduce communication and energy consumption

Multipath routing can overcome losses by duplicating and forwarding each sensor measurement

One or more other sensors have correctly overheard the packet

Some aggregate functions, such as SUM, COUNT, are duplicate-sensitive

Use RideSharing (RS) scheme for fault-tolerant, duplicate-sensitive aggregation

Page 27: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

27

RideSharing: Fault Tolerant Aggregation

Edges are classified into three types: primary, backup, and side edges

Using a small bit vector that each parent attaches to each data message it sends

Parents detect link errors when one or more children are missing from the bit vector

Figure3. Track Topology

Page 28: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

28

Cascaded RideSharing

Each parent broadcasts children ids and their bit positions inside its bit vector

When an error occurs, each backup parent

decides whether or not to correct the error based on its order in a correction sequence(parent with smallest id)

Page 29: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

29

References 

[1] Hai Liu, Amiya Nayak, and Ivan Stojmenovi ' Fault-Tolerant Algorithms/Protocols in Wireless Sensor Networks' Department of Computer Science, Hong Kong Baptist University, Springer-Verlag London Limited 2009

  [2] M.Yu, H.Mokhtar, and M.Merabti, 'A Survey on Fault

Management in Wireless Sensor Networks' School of Computing & Mathematical Science Liverpool John Moores University, 2007

[3] Farinaz Koushanfar1, Miodrag Potkonjak2, Alberto

Sangiovanni-Vincentelli1, ' FAULT TOLERANCE IN WIRELESS SENSOR NETWORKS'1Department of Electrical Engineering and Computer Science Univeristy of California, Berkeley , CA, US 94720, 2Department of Computer Science Univeristy of California, Los Angeles Los Angeles, CA, US 90095

  [4] Nithya Ramanathan, Kevin Chang, Rahul Kapur, Lewis Girod,

Eddie Kohler, and eborah Estrin,' Sympathy for the Sensor Network Debugger' UCLA Center for Embedded Network Sensing, ACM 2005

 

  

 

Page 30: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

30

References(cont’d)

[5] Jessica Staddon, Dirk Balfanz, Glenn Durfee' Efficient Tracing of Failed Nodes in Sensor Networks ', September 28, 2002, Atlanta, Georgia, USA, ACM.

[6] Sapon Tanachaiwiwat1, Pinalkumar Dave1, Rohan Bhindwale2, Ahmed Helmy1,' Secure Locations: Routing on Trust and Isolating Compromised Sensors in Location-Aware Sensor Networks ' 1. Department of Electrical Engineering – Systems 2. Department of Computer Science University of Southern California, ACM 2003

  [7] Gaurav Gupta1, Mohamed Younis2, ' Fault-Tolerant Clustering of

Wireless Sensor Networks ', Dept. of Computer Science and Elec. Eng. Dept. of Computer Science and Elec. Eng. University of Maryland Baltimore County University of Maryland Baltimore County 2003 IEEE  

Page 31: Elham.Hormozi@gmail.com & Rzh_asadi@yahoo.com Elham Hormozi & Razieh Asadi University of Science & Technology Mazandaran Babol 1

31

References(cont’d) [8] Jinran Chen, Shubha Kher, and Arun Somani,' Distributed Fault Detection

of Wireless Sensor Networks' Dependable Computing and Networking Lab Iowa State University Ames, Iowa 50010, 2006 IEEE

  [9] Sameh Gobriel, Sherif Khattab, Daniel Moss´e, Jos´e Brustoloni and Rami

Melhem,’ RideSharing: Fault Tolerant Aggregation in Sensor Networks Using Corrective Actions’, Computer Science Department, University of Pittsburgh,2006

  [10] Weiyi Zhang, Guoliang Xue and Satyajayant Misra,'Fault-Tolerant Relay

Node Placement in Wireless Sensor Networks', Department of Computer Science and Engineering at Arizona State University, IEEE INFOCOM 2007

  [11] S Harte1, A Rahman1, K M Razeeb2 'FAULT TOLERANCE IN SENSOR

NETWORKS USING SELF-DIAGNOSING SENSOR NODES', 1 University of Limerick, Ireland 2 Tyndall National Institute, Ireland,2005