ASCENT: Adaptive SelfConfiguring sEnsor Networks Topologies

1
ASCENT Adaptive Self-Configuring sEnsor Networks
Topologies

• Authors Alberto Cerpa, Deborah Estrin
• Presented by Suganthie Shanmugam

2
Presentation Topics

• Introduction
• Assumptions and Contributions
• ASCENT Design
• Analytical Performance Analysis
• Experimental Simulation
• Simulation Results
• Related Work
• Conclusion

3
Introduction

• Advances in micro-sensor and radio technology
• Smart sensors deployed in wireless network
• Nodes perform local processing
• Reduce communications and energy costs
• Low per-node cost ? densely distributed network
• Results in non-uniform communication density
• ASCENT
• Only a subset of nodes necessary to establish
  routing as node density increases
• Each node assesses its connectivity and
  adaptively self-configures to underlying topology

4
ASCENT Introduction

• How It works
• A node signals when it detects high packet loss
• Requests other nodes to join the network
• Reduces its load and does not join network till
  it is helpful to do so
• Adaptive configuration cannot be done from a
  central node
• Single node cannot sense conditions of nodes
  distributed in space
• Other nodes will be required to communicate
  detailed information to central node

5
Assumptions and Contributions

• Distributed Sensor Network Scenario
• Ex A habitat monitoring sensor network
• Sensors hand-placed or dropped from a plane
• Conditions
• Ad-hoc deployment
• Sensor network cannot be deployed in regular
  fashion
• Uniform deployment does not correspond to uniform
  connectivity
• Energy Constraints
• Expend minimal energy to maximize network
  lifetime
• Unattended operation under dynamics
• Preclude manual configuration and design-time
  pre-configuration

6
Assumptions and Contributions

• Easier to deploy large number of nodes initially
• Too few nodes used
• Distance between neighboring nodes large
• Packet loss rate increases
• Energy required to transmit prohibitive
• All nodes used
• Unnecessary energy expended
• Nodes interfere with each other channel
  congestion
• Perfect platform for ASCENT design

7
Assumptions and Contributions

• Assumption CSMA MAC protocol used in network
• Resource contention when many nodes involved in
  routing
• ASCENT
• Does not detect or repair network partitions
• Is not suitable when node density is low
• All nodes required to form effective network
• Two primary contributions
• Use of adaptive techniques to configure the
  underlying network
• Saves Energy, Extends Network lifetime
• Use of self-configuring techniques
• Reacts to operating conditions locally

8
ASCENT Design

• ASCENT adaptively elects Active nodes
• Awake all the time and perform multi-hop packet
  routing
• Passive nodes
• Periodically check if they should become active

9
ASCENT Design - State Transitions
10
ASCENT Design - Parameters Tuning

• NT (Neighbor Threshold)
• Average degree of connectivity in the network -
  Set to 4
• LT (Loss Threshold)
• Max. amount of data loss that an application can
  tolerate
• Application dependent Set to 20
• Tt, Tp Test Timer, Passive Timer
• Max. time a node remains in test and passive
  states
• Tt 2 minutes Tp 4 minutes
• Ts Sleep Timer
• Amount of time a node sleeps to conserve energy
• Large Ts Large energy savings but doesnt react
  to dynamics

11
Neighbor and Data Loss Determination

• Number of active neighbors, Avg. data loss rate
• Values measured locally by each node while in
  passive and test states
• Definitions
• Neighbor node - From which certain of packets
  received
• History Window CW Keep track of packets
  received from each node
• Each node increases the sequence number when each
  packet is transmitted
• When a sequence number is skipped, loss is
  detected
• Final packet loss
• Filter constant ? set to 0.3

12
Neighbor and Data Loss Determination

• The number of active neighbors (N)
• Number of neighbors with link packet loss smaller
  than the neighbor loss threshold (NLS)
• NLS 1- (1/N)
• N the number of neighbors calculated in the
  previous cycle
• If neighbor packet loss gt NLS, node deleted from
  list
• As number of neighbors increase, NLS should be
  increased
• Average data loss rate (DL)
• Calculated based on application data packets
• Detected using data sequence numbers
• If message not received from any neighbor - data
  loss
• Control messages are not considered
• Help, neighbor announcement and routing control

13
Interactions with Routing

• ASCENT
• runs above link and MAC layer below routing layer
• is not a routing or data dissemination protocol
• decides which nodes should join the routing
  infrastructure
• Nodes become active or passive independent of
  routing protocol
• Does not use state gathered by the routing
  protocol
• Does not require changing the routing state
• Test state (actively routing packets) ? passive
  state (listen-only)
• Cause some packet loss
• Improvement Traffic could be rerouted in
  advance by informing the routing protocol of
  ASCENTs state changes

14
Performance Analysis Goals and Metrics

• One-Hop Delivery Rate
• Measures of packets received by any node in
  network
• Indicates effective one-hop bandwidth available
  to nodes
• When all nodes are turned on Active case
  packet reception includes all nodes.
• ASCENT case - includes all except nodes in sleep
  state.
• End-to-End Delivery Rate
• Ratio of Number of distinct packets received by
  destination to the Number originally sent by
  source
• Provides an idea of quality of paths in the
  network and the effective multi-hop bandwidth

15
Performance Analysis Goals and Metrics

• Energy Savings
• Ratio of energy consumed by Active case to Energy
  consumed by the ASCENT case
• Average Per-Hop Latency
• Measures average delay in packet forwarding in a
  multi-hop network
• Provides estimate of end-to-end delay in packet
  forwarding

16
Analytical Performance Analysis

• Assumptions
• Nodes randomly distributed in an area A
• Average degree of connectivity (n)
• Packets propagated using flooding with random
  back-off
• Probability of successfully transmitting a packet
  
• P (success) (S 1)/ST
• Node density increase ? P (success) decreases
• When all nodes can transmit and receive, T n
• Since every node in vicinity can transmit
• Node density increase ? P (collisions) increases

17
Analytical Performance Analysis

• Average latency per hop related to S and T
• S No. of slots
• T No. of active nodes
• Each T node picks a random slot
• S1, S2ST
• Mean S / 2
• Uniform probability distribution

18
Analytical Performance Analysis
19
Analytical Performance Analysis

• P(d) distribution for different T and S 20
• T n
• When all nodes can transmit and receive
• As n ?, P(d) ?
• In ASCENT case
• T NT
• Independent of n
• P(d) remains constant

20
Analytical Performance Analysis

• Energy Savings
• Numerator Power consumed by all nodes without
  ASCENT
• Denominator Power consumed by all nodes running
  ASCENT
• 1 Power consumed by NT nodes selected by ASCENT
  to have their radios on
• 2 Energy of non-active nodes in passive state
• 3 Energy consumed in sleep state

21
Analytical Performance Analysis

• Energy Savings
• a Ratio of passive timer to sleep timer
• ß Ratio of sleep mode to idle mode power
  consumption
• NT fixed, ß small, as density ? power
  consumption is dominated by passive nodes
• When a small and Ts gtgtTp, large energy savings
• Large Ts ? slow reaction of passive nodes

22
Analytical Performance Analysis

• Energy savings of ASCENT with Adaptive timers
• No asymptotic behavior
• Energy savings increase linearly with density
• Slope of line primarily determined by Probability
  Threshold Pt

23
Simulation Experimental Methodology

• Implementation
• LinkStats module
• Adds increasing sequence number to each packet
• Monitors packets
• Maintains packets statistics
• Neighbor Discovery module
• Sends and receives Heartbeat messages
• Maintains list of active neighbors
• Energy Manager module
• Evaluate Energy Usage
• Acts as simulated battery

24
Simulation Experimental Methodology

• Simulator
• Built-in simulator (emsim) of EmStar used
• Provides channel simulator to model environment
  behavior
• Statistical model
• Experimental Test bed
• Total of 55 nodes used, All nodes wall-powered
• Routing
• Flooding used as routing protocol for simplicity
• On receiving a packet, flood module waits for a
  random time
• Randomization interval 5 seconds

25
Simulation Experimental Methodology

• Scenarios and Environment
• Experiments conducted with different densities
  ranging from 5 to 40 nodes
• Density defined topologically
• Defined by average degree of connectivity between
  all nodes not by physical location
• Achieved by adjusting transmit power of the RF
  transceiver
• Average number of hops 3
• Traffic
• One source sends approximately 200 messages
• Data Rate 3 messages / minute
• Nodes do not experience congestion

26
Simulation Results Network Capacity

• No major difference between analytical and
  simulated performance
• Active case
• All nodes join network and forward packets
• Low delivery rate
• As node density increases, P (collisions)
  increases
• ASCENT case
• Limits active nodes
• Channel contention does not increase

27
Simulation Results Network Capacity

• No. of hops 3
• Experiments
• No. of hops 6
• Simulations
• Increase in density
• ASCENT performs better than ACTIVE case
• Remains stable

28
Simulation Results Energy Savings

• ASCENT provides significant Energy savings
• As density increases
• Fixed State Timers
• Energy savings do not increase proportionally
• Number of Active nodes remains stable
• Adaptive State Timers
• Energy savings increase proportionally
• Passive nodes aggressive

29
Simulation Results Latency

• ACTIVE case
• As density ?, average per-hop latency is reduced
• Larger probability of a node picking a smaller
  random interval to forward the packet
• ASCENT
• As density ?, average per-hop latency remains
  stable
• Number of nodes able to forward packets remains
  constant

30
Results Reaction to Dynamics

• Evaluate how ASCENT reacts to node failures
• Let system run till stable topology reached
• Manually kill set of active nodes
• At high density, end-to-end delivery rate does
  not decrease
• High probability of a passive node to fix
  communication hole
• ASCENT with adaptive state timers more stable

31
Results Sensitivity to Parameters

• Larger randomization interval
• average one-hop delivery rate increases
• Increases end-to-end latency
• ASCENT outperforms ACTIVE case

32
Conclusions and Future Work

• Paper describes design, implementation, analysis,
  simulation and experimental evaluation of ASCENT
• ASCENT
• Has potential to significantly reduce packet loss
• Increases Energy efficiency
• Was responsive stable under varied conditions
• Future Work
• Evaluate interactions of ASCENT with MAC
• Investigate use of load balancing techniques
• Understand relationships between ASCENT and other
  routing strategies