A Faulttolerant SoftState Multihop Rendezvous Reservation Protocol for AdHoc Wireless Sensor Network

1
A Fault-tolerant Soft-State Multihop Rendezvous
Reservation Protocol for AdHoc Wireless Sensor
Networks

• Sripriya Vasudevan
• DAWN Research Group Meeting
• September 26 2000

2
Motivation

• Provides monitoring and control capability for a
  variety of applications
• Manufacturing
• Health Care
• Security
• Function independently or as part of embedded
  devices
• Power constrained operation
• Low power, low complexity devices
• Necessity to extend their life spans

3
Assumptions

• Power dissipation is dominated by
  beaconing/handshaking activities of the nodes
• Power dissipation due to user data transmission
  is negligible
• Conservation of power
• Limit the operations of nodes to a minimum

4
Network Architecture
Fat Node
Broadcast cluster
B
A
S
4
1
5
2
3
C
D
E
D
Lite Node
5
Fat Nodes (FaN)Lite Nodes(LiN)

• FaNs are
• High complexity, high power nodes
• Fewer in number
• Have long range beacons
• LiNs are
• Low complexity, low power devices
• Larger in number
• Have short range beacons

6
Communication Constraints

• LiN to LiN Only via multihops
• LiN to FaN Via Multihop
• (Except for nodes that are close to each other)
• FaN to LiN Via multihop or single hop
• FaN to FaN Multihop through the LiNs
• LiNs aid in data gathering and data forwarding

7
Other Assumptions

• Network is a hybrid of self-organizing/infrastruct
  ure wireless networks
• Nodes are stationary
• Channel conditions change very slowly
• Nodes have overlapping broadcast clusters
• Nodes support multiple applications
  simultaneously

8
Rendezvous Mechanism

• Operates between the routing layer above and the
  MAC layer below
• For power conservation,
• Nodes operate only when necessary and are in
  off state at other times called sleep mode
• Rendezvous The coordination mechanism that
  dictates these on and off times of nodes
• A periodic rendezvous is set for each neighbor of
  a node during which they exchange data

9
Concerns

• How does a node establish rendezvous instances
  (RIS) with its neighbors?
• What is the optimum rendezvous interval for
  minimal power consumption given a maximum
  response time (MRT)?
• How does selection of a particular RIS for the
  node pairs affect uniform expiration rate of all
  nodes?
• What are the implications of varying packet sizes
  on processing power?

10
Concerns (Contd.)

• What are the implications regarding efficiency
  and effectiveness of RIS mechanisms and routing
  protocols on each other?
• How do MAC protocols benefit by this mechanism?
• What are the operating parameters to achieve a
  desired degree of performance and power
  consumption?

11
Routing Protocol

• Gives the route between source and destination
  for all information exchange between both FaNs
  and LiNs
• MAC Protocol
• Allocates channel and prevents channel
  interference between sensor nodes

12
Definitions Notations

• Node inter-rendezvous period, tR Initially set
  to tRMAX seconds
• Maximum response time (MRT) is the QoS constraint
  
• Reservation Each session supporting a particular
  application
• Associated with each reservation is a timeout
  interval t0
• There is a limit on the number of reservations a
  system can support

13
RRP Algorithm

• Given a particular MRT value, the RRP makes a
  reservation along the source and destination
  nodes.
• As the RRP setup packet travels along the path
  from the source to the destination, the value of
  tR is appended to this packet
• As the RRP setup packet arrives at the
  destination, this node re-assigns the set of tRs
  required to satisfy the desired MRT

14
RRP Algorithm (Contd)

• Destination node will return an RRP-ACK to the
  source node along the same path
• Each node reads the assigned tR value and they
  update their rendezvous rate accordingly
• Once RRP arrives at the source, all node pairs
  along the route will be rendezvousing with an
  updated tR which will achieve the desired MRT

15
Example
Z
Y
X
tYZ
tXY
Initially tXY 50 tYZ 50 Desired MRT
70 Option 1 tXY 50 tYZ 20 Option 2 tXY
35 tYZ 35
Assignments NEVER INCREASE the current value of
a rendezvous interval ONLY DECREASES are allowed
16
Node Information

• Each node maintains a list of all reservations
  passing through it with the corresponding tR
  value
• Each node has info on the closest FaN
• Each reservation has a timeout interval T0
• After T0, if the reservation must continue
  indefinitely, then it must be refreshed by
  propagating a new setup packet along the route
• Rendezvous intervals may be updated to reflect
  new conditions
• If reservation times-out, it is removed from
  the list
• The interval may be increased to the next-lowest
  rendezvous interval on the list

17
Limitations

• Increases processing overhead for one node , the
  destination node
• Information overhead for destination node also
  increases
• Delay in assignments of tR values
• Setup packet size increases with the number of
  nodes along the route
• RRP- ACK packet is also large
• Multiple reservations and time outs might occur
  during setup ACK phase

18
Issues to be addressed

• Total delay ( MAC transmission time) must be
  less than or equal to MRT
• Total power dissipated in the system, which is
  proportional to the time and packet size
  processed must be minimal
• Maximize the number of reservations that the
  system can support

19
Heuristics

• Centralized Approach
• Destination based rendezvous assignments
  discussed earlier
• Distributed Approach

20
Heuristic 1

• Centralized approach (FaN closest to the
  destination)
• Follows the same RRP algorithm, except that the
  FaN closest to the destination does the
  rendezvous assignments
• Advantage FaN has more power and better handles
  the assignment overhead
• Disadvantage Delay would not be significantly
  different from the RRP
• RRP ACK packet size could be reduced if the
  assignments are sent individually for each node

21
Heuristic 2

• Distributed Approach (Among LiNs)
• Node pairs along the route reduce their tR values
  locally and recursively by some X units till the
  desired MRT is met
• Advantage One node is not burdened with all
  assignments
• Disadvantages
• Complex coordination among nodes
• Potential to reduce tR to a very low value which
  is undesirable for power savings

22
Heuristic 3

• Distributed Approach (Among FaNs)
• FaNs would need to have a global picture of the
  network connections reservations
• To do rendezvous assignments,
• LiNS can send their neighbor tables(NT) to the
  closest FaN
• FaNs can beacon the LiNs along the route to send
  the NT
• The FaNs do the assignments to the LiNs in a
  distributed fashion

23
Heuristic 3(Contd.)

• Advantages
• Distributed Does not drain one FaN
• LiNs conserve a lot of power as they are not
  involved
• Algorithm for the assignments can be complex,
  aiding in increasing the scalability and limit on
  the number of reservations
• Disadvantage
• Complex algorithm might be expensive to implement

24
Possible Failures

• Timing Faults
• Send messages that are late
• Omission Faults
• Omit messages
• Value Faults
• Send with erroneous content
• Send unspecified or impromptu messages

25
Possible Failures(Contd.)

• Crash Failures
• Halting Crash
• When nodes die down and are no longer part of the
  network
• Amnesia Crash
• When nodes die down and come back to life after a
  short time
• Design Failure
• By flaws in the algorithm
• Security Flaws
• When malicious nodes are involved in some or all
  of the above failures (except design)

26
System Paramaters

• Number of nodes Both LiNs and FaNs necessary for
  the operation
• Bandwidth Type of applications a system can
  support and scalability for the same
• Packet sizes Varying packet sizes and their
  effect on power consumption of the network nodes
• Power Levels Network behavior depending on
  varied power levels of operation
• Traffic Pattern Has to be varied to obtain
  system response

27
Performance Analysis

• Overhead transmission and power
• Set-up time and Delay
• Power consumed per packet, per LiN, per FaN
• Rate of death of nodes in the network