Fault Management in Mobile Ad-Hoc Networks

by Tridib Mukherjee

Transient Faults in Mobile Ad-Hoc Networks

Mobility of the NodesError Prone MediumLink FailuresLow Battery PowerNode Corruption

Fault Tolerance

Ability of a system to perform its function correctly even in the presence of internal faults.

Makes the network system more dependable.

Hides the faults from the user.Two basic kinds of Fault Tolerance :

Proactive and Reactive.

Self Stabilization

Stabilizes a Distributed System to a legitimate state from any arbitrary initial state.

Used as a Proactive Fault Tolerant Scheme.

There are 2 properties : Closure and Convergence.

Self-stabilizing Multicast Routing Protocols For Mobile Ad-Hoc Networks

Shortest Path Spanning Tree (SPST)Used in this project

Maintains Shortest Path from source to destination.

Beacon Messages provide information about neighbors.

Recreates the tree in case of faults.Unnecessarily propagates limited faults

across the network.

Fault Containment

Contains the fault in the region where it has occurred

Improves stabilization time Considerably.Increases Computational and

Communication Overhead.Does not contain faults in all the desired

cases in Mobile Ad-Hoc Networks.Tradeoff needed for optimal Energy

Efficiency while managing the faults.

Fault-containment Algo.

can_stabilize :

Propose for Adaptation

Adapt to the changing Fault Scenarios.Use Self Stabilization where Fault

Containment can not contain the faults.Use Fault Containment where it can

contain the faults.

Fault Classification

Fault-Containable (FC) Faults Fault can be contained using Fault Containment

Non-Fault-Containable (NFC) Faults Fault Containment can not contain the faults Self-stabilization and Fault-containment have same

performance Fault-containment executes self-stabilization

internally Fault-containment adds computational overhead

Valid SPST Tree

R

A

X

M

Y

Level 0

Level 1

Level 2

Level 3

Level 4

M moves out

R

A

C

X

D

M

Y

Level 0

Level 1

Level 2

Level 3

Level 4

can_stabilize(Y) is falsecan_stabilize(X) is falsecan_stabilize(C) is falsecan_stabilize(D) is false

This is NFC fault

R

A

X

Y

Level 2

Level 3

Neighborhood of Y is different

R

A

C

X

D

M

Y

Level 0

Level 1

Level 2

Level 3

Level 4

can_stabilize(Y) is truecan_stabilize(A) is falsecan_stabilize(C) is falsecan_stabilize(D) is false

FC faults

The scenario of the previous slideFaults occurred due to corruption are FC

faultsBoth NFC and FC faults can occur in

multiple nodes simultaneouslyFor NFC faults, self stabilization is

executed internally

Two FC faults within 2 hops

R M

X

AY B

• Level of A gets corrupted to 6 • M moves out and X becomes the parent of Y • Both the FC faults become Non-containable • Distance of 2 hops is named as Containability Limit (CL)

0 1 2 3 4 5

Reason

Gp(Y) and Gp(A) are true

Can_stabilize(Y) and Can_stabilize(A) are false

So Fault Containing Algorithm executes self-stabilization internally

FC faults becomes NFC if they occur within CL

Improved_Can_Stabilize

Check if local action can nullify Gp in all the two hop neighbors instead of one hop neighbors as in the original algorithm

Otherwise check if local actions in all the one hop neighbors can eradicate fault in all the two hop neighbors

Improved Fault-containment

Reasoning

Gp(Y) and Gp(A) are true

Can_stabilize(Y) and Can_stabilize(A) are also true

Fault Containing Algorithm executes self-stabilization internally only if faults are NFC

Containability Limit is 0

Simulation

Simulation is done in NS2Comparison between Self-stabilization,

Fault-containment and Improved Fault-containment

Simulation is done for NFC and FC faults as well as multiple FC faults occurring within CL

Performance is measured in terms of Beacon Intervals

NFC Fault Simulation Result

FC faults with distance greater than CL

FC faults with distance less than CL

Advantages & Disadvantages

If a fault can be contained, it is contained regardless of its occurrence in the network

Costs more communication overhead if a fault is not containable

References

Sukumar Ghosh, Arobinda Gupta, Sriram V. Pemmaraju. ”Fault-containing network protocols”. Proceedings of the 1997 ACM symposium on Applied computing, p.431-37, April 1997, San Jose, California, United States.

Sukumar Ghosh, Arobinda Gupta, T. Herman, Sriram V. Pemmaraju. Faultcontaining Self-Stabilizing Algorithms”. 15th Annual ACM Symposium on Principles of Distributed Computing, 1996, pp. 45-54.

References (Contd…)

Sandeep K. S. Gupta, Pradip K. Srimani. ”Self-stabilizing multicast protocols for ad hoc networks”. Journal of Parallel and Distributed Computing 63(1): 87-96 (2003)

Questions?