Self Healing (2)

22
Wireless Sensor Networks Self-Healing Professor Jack Stankovic University of Virginia 2005

Transcript of Self Healing (2)

Page 1: Self Healing (2)

Wireless Sensor Networks

Self-Healing

Professor Jack Stankovic

University of Virginia2005

Page 2: Self Healing (2)

WSN Property

• Unattended long-term operation

– System Initialization• Set system parameters

– Adapt to Wireless communication changes

– Adapt to environmental changes • Impact on sensor thresholds, etc.

– Adapt to failures– Adapt to workload changes– Adapt to power reductions

Page 3: Self Healing (2)

Self-Healing• One aspect of self-* system

– Needs self-monitoring

– Self-organizing– Self-managing– Self-calibration (turn-key system?)– Self-power management– …

– Called Autonomic Computing

Page 4: Self Healing (2)

Self-Healing• Relates to Fault Tolerance

– Mask (correct) an error• ECC• Are you healed with respect to that error?

– Yes for that error, but maybe not the cause!– Detect an error

• Lost message• Re-send message

– Ignore error (design for it)• Robustness - Act in presence of error (are

you healed)?• Example: WSN still works if 20% of nodes

are dead

Page 5: Self Healing (2)

Self-Healing •Detect and heal

– Example: Detect neighbor is “dead”, try restarting it and if successful -> healed

•From a system perspective– Wide spectrum of capabilities– Not binary

Page 6: Self Healing (2)

Self-Healing

• In Localization– Heal: If node fails to obtain location

during walking GPS, it gets info from neighbors and uses tri-lateration

– Heal: A node detects that it lost its localization information, then re-determines it from its neighbors

Page 7: Self Healing (2)

Recall APIT Algorithm

• Assumption: An area covered with heterogeneous nodes. – Anchor nodes equipped

with high-powered transmitter.

– Location information obtained from GPS.

• Location estimation by Area-based Approach.• Narrow down the location of

one node by deciding its presence inside or outside the triangles formed by the anchors.

Estimated Location

A

Green-Anchors

Example: 14 anchors, butThere are 100s of nodes like A

Page 8: Self Healing (2)

Self-Healing

• In Routing– Multiple parents in backbone tree

• If detect one parent is dead, use the other• Local decision on choosing alternative

parent is fast• Heal: Re-create n-parent tree on system

rotation• In MAC

– Retransmit lost packet– Heal: Enough lost packets -> change

power level or other thresholds

Page 9: Self Healing (2)

Self-Healing

•In System Initialization–Each phase is coordinated and

sequential– If a node is not in-step it becomes silent–Heal: Silent node (possibly) becomes

active at next system rotation

Page 10: Self Healing (2)

Time-Driven System Operation

RESET

Phase ISystem Initialization

Phase IIILocalization

Phase VNetwork Partition & Diffusion

Tree Constrcution

Phase VISentry Selection

Phase VIIHealth Report

StartPhase VIII

Power Mgmt

Event Tracking

Phase IITime Sync

Phase IVAsymmetri Detection

Phase VIII

Event Tracking

Power Mgmt

Dormant Section

Tripwire Section

Wakeup Service

Page 11: Self Healing (2)

Self-Healing• In Wakeup

– Decentralized and if some nodes fail to wake-up it is not a problem because many others will be awake

– Heal: If it keeps happening then perhaps increase wakeup preamble and if this works then the wakeup has been healed

Page 12: Self Healing (2)

Duty Cycle

W

W

Preamble

Page 13: Self Healing (2)

Self-Healing

•In Sensing–Fail-stop – use of many sensors in WSN

and since targets move, problems are masked (robust)

–Heal: Byzantine failure – detect that a node is continuously reporting and shut it down

•In Tracking– If group leader fails–Heal: Info is still with the group members

and is passed to next leader

Page 14: Self Healing (2)

Group Management (Tracking)

Base Station

Page 15: Self Healing (2)

Underlying Self-Healing in WSN

•Limited Effect– Clock sync, neighbor discovery, etc. are highly

decentralized and local. Single node failures (hopefully) only affect that node and do not propagate to the rest of the network.

Page 16: Self Healing (2)

Self-Healing Mechanism

• System Rotation– Can correct many issues – Can be executed based on time– Could be extended to re-run when

many failures are detected, BUT this means extra detection messages which affects lifetime and stealthiness!

Page 17: Self Healing (2)

Security in WSN

• Solutions from the start• System must operate in presence of faults

AND attacks• Framework for security updates as attacks

evolve over time

• Solution: Adaptable Self-Healing as Security Support in Wireless Sensor Networks

Page 18: Self Healing (2)

Confluence of Techniques

• Self-Healing• Aspects

• Decentralized control with diversity• Wireless downloads

Page 19: Self Healing (2)

Adaptive Self-Healing/ Aspects

Partition

Routing

AdvicePoint Cuts

Download New Entities

Component-Based WSN

Self-Healing

MONITOR

Page 20: Self Healing (2)

Decentralized Control

• Redundancy• Mask faults• Uniformity a problem

– Diversity

Page 21: Self Healing (2)

Wireless Download

• New executables (or WSN individual components)– Includes monitoring

• New Join Points and Point Cuts• New Advice

Page 22: Self Healing (2)

Summary

• Unattended operation over long lifetime– Require self-*

• Good software design and implementation

• Good FT techniques• Security attack models and healing

– Denial of Service– Mis-information (may be of more use in

some situations)