Monitoring and Debugging of Deployed Sensor Networks Matthias - - PowerPoint PPT Presentation

monitoring and debugging of deployed sensor networks
SMART_READER_LITE
LIVE PREVIEW

Monitoring and Debugging of Deployed Sensor Networks Matthias - - PowerPoint PPT Presentation

Monitoring and Debugging of Deployed Sensor Networks Matthias Ringwald, Kay Rmer, ETH Zrich Overview Wireless Sensor Network (WSN) Deployment State of the Art in WSN Debugging Non-Intrusive Debugging of WSNs Summary &


slide-1
SLIDE 1

Monitoring and Debugging of Deployed Sensor Networks

Matthias Ringwald, Kay Römer, ETH Zürich

slide-2
SLIDE 2

2

  • 12. Okt. 2005

Matthias Ringwald, Kay Römer, Institut for Pervasive Computing

Overview

  • Wireless Sensor Network (WSN) Deployment
  • State of the Art in WSN Debugging
  • Non-Intrusive Debugging of WSNs
  • Summary & Discussion
slide-3
SLIDE 3

3

  • 12. Okt. 2005

Matthias Ringwald, Kay Römer, Institut for Pervasive Computing

Network of many sensor nodes:

  • battery powered
  • wireless
  • environmental sensors

Environmental monitoring

  • Great Duck Island
  • Glacier
  • Volcano

BTnode Rev3

Wireless Sensor Networks (WSNs)

slide-4
SLIDE 4

4

  • 12. Okt. 2005

Matthias Ringwald, Kay Römer, Institut for Pervasive Computing

WSN Life Cycle

Unexpected problems in deployed networks

  • Moisture
  • Depleted batteries
  • Incorrect sensor readings
  • Frequent connectivity changes

Simulation, emulation Lab testbed Configuration & test Operation

Deployment

slide-5
SLIDE 5

5

  • 12. Okt. 2005

Matthias Ringwald, Kay Römer, Institut for Pervasive Computing

  • Reasons
  • Simulation: does not capture all real-world effects
  • Testbed: few nodes, controlled environment
  • Users: Application experts, not system experts

⇒ Problems are inevitable and cannot avoided ⇒ Get insight into deployed sensor network

Simulation, emulation Lab testbed

Configuration & test

Operation

“Unexpected” Problems?

slide-6
SLIDE 6

6

  • 12. Okt. 2005

Matthias Ringwald, Kay Römer, Institut for Pervasive Computing

State of the Art

Wireless network used for data gathering

  • Nucleus: attribute query, event logging [Tolle2005]
  • Sympathy: In-network state collection, statistic analysis

[Ramanathan2005]

Additional code/serial communication

  • Deployment support network (DSN): additional, temporary

wireless network as backbone [Beutel2005]

Non-intrusive to deployed sensor node

  • TinyOS message center: gateway packet monitor [2003]

intrusiveness

slide-7
SLIDE 7

7

  • 12. Okt. 2005

Matthias Ringwald, Kay Römer, Institut for Pervasive Computing

BTnut on BTnode3

  • Cooperative Multithreaded OS

+ POSIX: Signals, Mutex, Semaphore, Device Driver…

  • 260 KiBi RAM
  • Bluetooth Radio
  • Chipcon CC1000

Deployment Support Network Deployment Support Network

More Info: btnode.ethz.ch

slide-8
SLIDE 8

8

  • 12. Okt. 2005

Matthias Ringwald, Kay Römer, Institut for Pervasive Computing

Non-intrusive debugging of WSNs

Mission statement: “Monitor & debug WSN by non-intrusive observation of radio communication”

Sniffer Decoder Evaluation

slide-9
SLIDE 9

9

  • 12. Okt. 2005

Matthias Ringwald, Kay Römer, Institut for Pervasive Computing

Distributed Sniffer

Generic MAC listening

  • Frequency, data rate
  • Start-Of-Packet symbol
  • Packet size

DB

SINGLE NODE SNIFFER IMPLEMENTED NEXT: INTEGRATION IN DSN

slide-10
SLIDE 10

10

  • 12. Okt. 2005

Matthias Ringwald, Kay Römer, Institut for Pervasive Computing

Generic Packet Decoding

  • Existing Solutions:
  • Top-Down: ASN.1 packet description
  • Bottom-Up: Ethereal network analyzer: decode code written

in C

  • New: “Attributed C Structs”
  • most WSN apps are developed in C
  • Structs are easy to understand/parse
  • Extensions allow for encapsulated

packets

IMPLEMENTED

/** TinyOS message */ struct TOS_Msg { uint16_t addr; uint8_t type; uint8_t group; uint8_t length; int8_t data [29 ]; uint16_t crc; };

slide-11
SLIDE 11

11

  • 12. Okt. 2005

Matthias Ringwald, Kay Römer, Institut for Pervasive Computing

Data Analysis

  • Topology Visualization: link-layer header info
  • Statistics : re-implement Sympathy?
  • Rule engine : assertions, failure tests

… eliminate bugs

PLANNING

slide-12
SLIDE 12

12

  • 12. Okt. 2005

Matthias Ringwald, Kay Römer, Institut for Pervasive Computing

Summary

  • Inevitable problems during WSN deployment
  • New tools needed to aid in debugging
  • Current approaches alter behavior of nodes
  • Our approach is non-intrusive, promising and can be

added on the spot

slide-13
SLIDE 13

13

  • 12. Okt. 2005

Matthias Ringwald, Kay Römer, Institut for Pervasive Computing

Discussion

  • Pro:
  • non-intrusive
  • access to all

communication

  • allows to understand/

correct communication issues

  • Contra:
  • multiple frequencies,

frequency hopping: hard

  • Internal state can only be

inferred

  • Extension:
  • nodes could send

additional/intrusive debugging data

Thanks!

slide-14
SLIDE 14

14

  • 12. Okt. 2005

Matthias Ringwald, Kay Römer, Institut for Pervasive Computing

Example Packet Description

struct Packet { u_int16 src; u_int16 dest; u_int8 type; u_int8 size; u_int8 data[size]; u_int16 CRC; }; struct RoutingPacket : Packet.data ( type == 1) { u_int8 nrHops; u_int16 nextHop; };