Dynamic Data Compression in Multi-hop Wireless Networks

Document Sample
Dynamic Data Compression in Multi-hop Wireless Networks Powered By Docstoc
					Dynamic Data Compression
in Multi-hop Wireless Networks

    Abhishek B. Sharma (USC)
    Collaborators:
    Leana Golubchik
    Ramesh Govindan
    Michael J. Neely
Data collection application in sensor networks

   Sensor nodes collect measurements
    that must be delivered at a sink.
   Multi-hop routing over a tree.
       Radios have limited transmission
        range
   Energy constrained
     Nodes are battery powered.




                       SIGMETRICS/Performance'09   2
Wireless sensor network platforms:
Radio is the energy hog

# CPU cycles for same
energy as 1 byte
transmitted
Processor: MSP430


                Data transmission is expensive.


            Transmission range:          increases

                                  Sensor network radios


Figure from Sadler and Martonosi (SenSys 2006)            3
Energy efficient data collection applications

   Need to transmit data using small energy budget.
   Challenge: Transmission costs lots of energy.
       Data is transmitted across multiple hops.
   Solution: Send less.
       compress data before transmitting energy trade-off.
         Transmission vs. Compression
   Energy cost of compression.
       Not just CPU computations.
       Memory access, FLASH access




                        SIGMETRICS/Performance'09             4
Data compression:
Exploring the energy trade-off
   Related work:
       Single vs. multi-hop routing (Sadler et al., SenSys’06).
       Evaluating the energy efficiency of various algorithms.
        (Barr et al., MobiSys’03).
       Designing “light” yet energy efficient compression algorithms
        (Sadler et al., SenSys’06).
   Sadler et. al., SenSys’06
       Single-hop: data compression does not save energy
       Multi-hop: data compression saves energy.
       “always compress” is not optimal.
   Energy trade-off was not explored in a “dynamic”
    environment.
                        SIGMETRICS/Performance'09                   5
   System dynamics
                                                                              Sink



  B            A          Sink          B            A       Sink
                                                                          A             B



Energy                       Energy                           Energy




      w/o comp. comp.                     w/o comp. comp.           w/o comp. comp.
         Don’t compress                       Compress                 Don’t compress

 System dynamics impact the energy savings from compression.

                                 SIGMETRICS/Performance'09                              6
Compression decision in a dynamic environment


   Compression decision: “When to compress?”
   Must adapt to system dynamics.
    1.   Network dynamics: Link quality, topology
    2.   Application-level: sampling rate
    3.   Platform upgrade: low power radios, compression
         algorithm
    “When to compress” is not straight forward to
     determine.
        “Always compress” policy may not work well.




                       SIGMETRICS/Performance'09           7
Data compression in a dynamic environment:
Stochastic Network Optimization
   The application data arrival process and time varying link
    qualities are modeled as ergodic stochatic processes.

   Goal: Minimize the total system energy expenditure.
       System energy expenditure: total energy expenditure
        across all the nodes.


   Constraint: Network is “stable”
       bounded average queue size at all the nodes.
       implies finite delay in delivering data to the sink.


                         SIGMETRICS/Performance'09             8
  Stochastic Network Optimization:
  Lyapunov Optimization technique1
 Arrival process                                     “Backpressure” based
                        Lyapunov drift
                           analysis                  transmission decisions
 Link dynamics


                                  Stability
                                                     “Backpressure” based
Lyapunov Optimization:                               transmission decisions
 Arrival process
                           Lyapunov drift
                              analysis +                     joint decision
 Link dynamics          Utility (energy cost)
 Compression
 at the source                         Energy-      Compression decision
                      Stability
                                       efficient    algorithm

1Georgiadis, Neely and Tassiulas. Resource Allocation and Cross Layer Control
in Wireless Networks, Foundations and Trends in Networking.
“Joint” compression and transmission decisions


                 Data transfer rate
 Compression                               Transmission
               Lots of retransmissions
   Decision     Application data rate        Decision
  Algorithm                                  Algorithm




               SIGMETRICS/Performance'09                  10
Our contributions
1.       Stochastic network optimization formulation
          First to consider data compression for multi-hop networks
           in a dynamic environment.
2.       Derive a “joint” congestion and transmission decision
         algorithm.

3.       Prove stability and analytical performance bounds.

4.       Propose and evaluate a distributed version.
          Works with CSMA MACs: 802.11, 802.15.4



                         SIGMETRICS/Performance'09                 11
   SEEC: Stable and Energy Efficient Compression
   System Model
                                                    Maintains a table of
                                                    avg. compression ratio and
                    Compression Module              avg. energy cost
 Application Data                                   for each comp. option k.




 Data from other
 nodes                                             Ul [t] = Un[t] - Um[t]
                     Un[t]                                                     m
Un[t]:              Transmission Module
                                                   l[t] = C(link quality, trans. power)
Queue backlog
                          Node n
  Decisions (every time slot t):
  Compression decision: whether to compress ? which option?
  Transmission decision: which nodes should transmit data?
                       SIGMETRICS/Performance'09                                   12
 SEEC: Transmission schedule
 “Queue differential backlog” based
                                                       Transmission rate
    Each link is assigned a weight.
                                                                       Transmit power


               Differential backlog       Control parameter

Scheduling
constraints                                                       Positive weight
                             Transmission                         links on which
  Link                         scheduler                          data transfer is
 weights                                                          allowed
    Negative weight on a link
        Either due to a small queue backlog or poor link quality

                           SIGMETRICS/Performance'09                                 13
Transmission decision:
Impact on queue backlog
   A node does not get to transmit till its backlog is greater
    than transmission threshold [t] = O (V/ [t]).
       Weight on its outgoing link should be positive.
   Increasing V results in higher queue backlog.
       Higher delay in delivering data to the sink.


   Avg. queue backlog grows will hop-count distance from
    the sink.


                                                    Sink


                        SIGMETRICS/Performance'09                 14
Compression decision:
Driven by queue backlog
   A node compresses data only when its queue backlog is
    greater than compression threshold [t].
       Directly proportional to compression energy cost.
       Inversely proportional to the average compression ratio.
       Increases as we increase the V.
   SEEC does not compute these thresholds explicitly.




                        SIGMETRICS/Performance'09                  15
   Example: SEEC in action
                       B                   A           Sink

        Transmit power = P (fixed)
        Link quality: “Good”= 2 Mbps, “Bad” = 1 Mbps                  B[t]
         Link links are sink becomes “Bad”
         Both from A to “Good”

A[t]
                                        Queue
                                        backlog                          B[t]

 A[t]                      Node B starts
                             No compression
                            compressing data

              Node A        time                              Node B       time

                           SIGMETRICS/Performance'09                           16
     SEEC’s Performance:
     Energy vs. Delay trade-off
                                         Theorem:




P*

     V (control parameter)



                         SIGMETRICS/Performance'09   17
Distributed version:
Implementing SEEC’s transmission decision
    Finding the optimum transmission schedule is NP-
     complete.
        Approximation algorithms are known.
1.    Global vs. Local information.
2.    802.11, 802.15.4 MACs:
      CSMA based (no timeslots).

    Positive queue differential heuristic (Sridharan et al.)
      Contend if (outgoing) link weight is positive

      Distributed version: dSEEC.




                       SIGMETRICS/Performance'09                18
Evaluation using Simulations

   Determining the model parameters
       Compression ratio and energy cost, transmission energy cost
   Measurements on real hardware: LEAP2
   Radio: 802.11b
   Compressed real-world sensor data from a bridge
    vibrations monitoring deployment (Paek et al.’ 06).
   Compression algorithm: zlib compression libraries.
    Simulator: Qualnet




                         SIGMETRICS/Performance'09                    19
     dSEEC: Summary of simulation results.

1.    10-30% energy savings compared to “always compress”.
        Tree-topology impacts the savings.




                          SIGMETRICS/Performance'09      20
 Compare with “Always compress”
                                        30 % reduction




  Cluster-Tree topology1
                                      Never        dSEEC      Always
Periodic application data arrival     compress                compress

Link quality did not change.

 1Used   in several deployments: Paek (WCSCM’06), Hicks (ImageSense’08)
     dSEEC: Summary of simulation results.

1.    10-30% energy savings compared to “always compress”.
         Tree-topology impacts the savings.


2.       Able to adapt to system dynamics.

3.       Sensitivity of energy savings to V


              Lots of simulation results in the paper


                           SIGMETRICS/Performance'09     22
     Conclusion
1.    Derived an algorithm for making compression decisions
      that is stable, energy-efficient, and adapts to system
      dynamics.
        Our work is the first to propose an adaptive algorithm for
         the multi-hop networks.
2.    Energy vs. Delay trade-off
        Proved Analytical bounds
3.    dSEEC: distributed version suited for CSMA MACs
4.    Significant energy savings compared to simple policies.
     Future direction:
        Consider in-network data aggregation and compression.



                         SIGMETRICS/Performance'09                    23
Algorithm derivation; proofs available in technical report.
             http://enl.usc.edu/~abhishek
                    Questions?




                  SIGMETRICS/Performance'09                   24

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:15
posted:9/11/2012
language:English
pages:24