INFOCOM Paper Review

1
INFOCOM Paper Review

• On-demand Power Management for
• Ad Hoc Networks
• By
• Rong Zheng and Robin Kravets
• Department of Computer Science
• University of Illinois

Review by Kaushik Sampath
2
Motivation

• Transmit power control consequences
• - Battery Power a constrained resource
• - Traffic Capacity
• Energy Conservation
• a. Not reflective of active communication but
  more on
• topological/geographical information.
• b. No significant difference between idle state
  than in
• S/R state
• Problems in Ad Hoc Networks
• a. Nodes can be data source/sink or routers.
• b. No centralized entity like an Access Point.

3
Ideas Proposed

• On demand power management for generic ad hoc
  networks that is not routing specific or MAC
  protocol specific (Integration of new protocols
  possible)
• Reduction of energy consumption while maintaining
  effective communication
• Integration of routing information and power
  management from MAC protocols
• Judiciously turn on/off radio based on active
  communication in the network
• Connection maintained only between pairs of
  senders and receivers and along route of data
  communication

4
Main Ideas

• Establish a soft-state timer for transitions
  between power management modes for each node
• Timer refreshed by data and control messages in
  the network
• Length of timer reflects adaptive nature of
  framework to varieties in traffic load
• Neighbor discovery necessary (Hello messages or
  snooping transmissions)
• Advantage Data delivery expedited without delay
  of waking up sleeping nodes along the route

5
Design Space

• Power management (PM) spans all layers
• Each layer has access to different type of
  information and has different mechanisms
• MAC layer PM using local info
• Network layer More global taking
  topology and traffic into consideration
• Scheme of framework scheme ranges from proactive
  to reactive with
• Proactive Always on
• Reactive Always off

6
Design Space of Power Management Schemes
7
Other protocols discussed

• MAC layer approaches
• - Time scale for PM can be per packet or shorter
  
• - Lack of topology/traffic information
• PAMAS Turns off radio when it overhears packet
  not addressed to it
• STEM
• - Similar to IEEE 802.11 but uses an independent
  
• control channel to avoid clock
  synchronization
• - Uses asynchronous beacon packets in a second
• control channel to wake up intended
  receivers
• S-MAC Modification of NAV for virtual channel
  reservation
• Pure MAC layer approach of IEEE 802.11 (most
  reactive)

8
Technical Details

• I. Cross layer design for Power Management
• Leverage both network layer and MAC layer
  information
• Knowledge about route setup and package
  forwarding when PM should be performed
• Route discovery (routing protocol) nodes
  along this route should be as responsive as
  possible
• PM protocol wake up sleeping receiver
  taken care of by MAC
• Integrate route discovery and MAC service

9
Technical Details (contd)

• II. Power Management Mode and State Transition
• Active mode node awake and ready to receive
• Power Save mode sleeping and waking up
  periodically to check for messages
• Transitions
• - Timer on for order of network life (proactive)
• - Timer zero throughout (reactive)
• Data messages and some control messages are good
  indicators

10
Technical Details (contd)

• III. Neighbors Power Management Mode
• Hello messages (at periodic time intervals)
  costly
• Passive inference (snooping in promiscuous mode)
• Challenges of Passive inference
• - Nodes in PSM cannot hear from neighbor
• - Nodes in PSM may not be transmitting
• Two indicators used by framework
• - Lack of communication during time interval
• - Packet delivery failure to the neighbor
• Pros and Cons of Passive inference
• - Does not rely on additional control messages
  and
• saves energy
• - Ambiguity of link failure and PM mode of
  neighbor

11
Comparison of Protocols

• Overview of IEEE 802.11 Power Management in
• Ad Hoc networks
• Authors prototype (state transition diagram)

12
Performance Evaluation

• Implementation parameters
• RTRQ_KEEPALIVE 0
• RTRL_KEEPALIVE 5
• DATA_KEEPALIVE 2
• SRC_KEEPALIVE 2
• DST_KEEPALIVE 2
• REFRESH_INTERVAL 5
• energy_goodput total_bit_transmitted
• total_energy_consumed

13
Performance Evaluation (contd)

• Effectiveness in terms of
• - Longevity
• - Efficiency
• Long lived CBR traffic in static networks (see
  figure below)

14
Performance Evaluation (contd)

• On-off traffic (see figure below)

15
Performance Evaluation (contd)

• Impact of mobility (see figure below)

16
Conclusions

• On demand power management framework that reduces
  energy consumption in ad hoc networks while
  maintaining effective throughput
• Explores design space between proactive and
  reactive management approaches by adapting to the
  traffic load in ad hoc networks

Future Work

• Online adaptation of keep alive timers when their
  values are not fixed
• Better handling of mobility
• Load Balancing

17
References

• On Demand Power Management for Ad Hoc
• Networks Rong Zheng, Robin Kravets
  University of
• Illinois at Urbana-Champaign
• http//lion.cs.uiuc.edu/zheng4/
• Wireless LANs Jim Geier, Sams 2nd Edition