Docstoc

Wireless Ad Hoc and Sensor Networks

Document Sample
Wireless Ad Hoc and Sensor Networks Powered By Docstoc
					Quorum-Based Asynchronous
  Power-Saving Protocols for
IEEE 802.11 Ad Hoc Networks
            Presented by

          Jehn-Ruey Jiang

 Department of Computer Science and
       Information Engineering
    National Central University
To Rest, to Go Far!
               Outline
 IEEE 802.11 Ad hoc Network
 Power Saving Problem
 Asynchronous Quorum-based PS Protocols
 Optimal Asyn. Quorum-Based PS Protocols
 Analysis and Simulation
 Conclusion
               Outline
 IEEE 802.11 Ad hoc Network
 Power Saving Problem
 Asynchronous Quorum-based PS Protocols
 Optimal Asyn. Quorum-Based PS Protocols
 Analysis and Simulation
 Conclusion
       IEEE 802.11 Overview
 Approved   by IEEE in 1997

 Extensions   approved in 1999

 Standard   for Wireless Local Area Networks
 ( WLAN )
     IEEE 802.11 Family(1/2)
 802.11a:
 6 to 54 Mbps in the 5 GHz band

 802.11b (WiFi, Wireless Fidelity):
 5.5 and 11 Mbps in the 2.4 GHz band

 802.11g:
 54 Mbps in the 2.4 GHz band
        IEEE 802.11 Family(2/2)
 802.11c:   support for 802.11 frames
 802.11d:   new support for 802.11 frames
 802.11e:   QoS enhancement in MAC
 802.11f:   Inter Access Point Protocol
 802.11h:   channel selection and power control
 802.11i:   security enhancement in MAC
 802.11j:   5 GHz globalization
         IEEE 802.11 Market
                         Source: Cahners In-Stat

($ Million)
Infrastructure vs Ad-hoc Modes
  infrastructure
   network

                            AP

               AP         wired network
                                               AP




                    Multi-hop ad hoc network




 ad-hoc network                                     ad-hoc network
   Ad hoc Network Applications

 Battlefields

 Disaster   rescue

 Spontaneous     meetings

 Outdoor    activities
               Outline
 IEEE 802.11 Ad hoc Network
 Power Saving Problem
 Asynchronous Quorum-based PS Protocols
 Optimal Asyn. Quorum-Based PS Protocols
 Analysis and Simulation
 Conclusion
                Power Saving
 Battery   is a limited resource for portable
  devices
 Battery technology does not progress fast
  enough
 Power saving becomes a critical issue in
  MANETs, in which devices are all supported
  by batteries
    Solutions to Power Saving
 PHY   Layer: transmission power control
   Huang (ICCCN’01), Ramanathan (INFOCOM’00)


 MAC   Layer: power mode management
   Tseng (INFOCOM’02), Chiasserini (WCNC’00)


 Network   Layer: power-aware routing
   Singh (ICMCN’98), Ryu (ICC’00)
     Transmission Power Control
 Tuning transmission energy for higher channel
  reuse
 Example:
     A is sending to B (based on IEEE 802.11)
     Can (C, D) and (E, F) join?       No!
                                        Yes!


                          B
           C
                D
                     A        F     E
       Power Mode Management
 doze mode vs. active mode
 Example:
     A is sending to B
     Does C need to stay awake? No!
                                    It can turn off its radio
                              B     to save energy!
                     A            But it should turn on its
                          C
                                  radio periodiclally for
                                  possible data comm.
          Power-Aware Routing
 Routing in an ad hoc network with energy-
  saving (prolonging network lifetime) in mind
 Example:


                N1     +          N2   +
    SRC                –               –             DES
                                                      T
           +               Better!!              +
           –                                     –
                                           +
                            +
                 N3               N4       –
                            –
                Our Focus
 Among   the three solutions:
  PHY Layer: transmission power control
  MAC Layer: power mode management
  Network Layer: power-aware routing
 IEEE 802.11 PS Mode(2/2)
Environments:
 Infrastructure (O)


 Ad hoc (infrastructureless)
   Single-hop (O)
   Multi-hop
      IEEE 802.11 PS Mode(1/2)
An  IEEE 802.11 Card is allowed to turn off its
radio to be in the PS mode to save energy
Power Consumption:
(ORiNOCO IEEE 802.11b PC Gold Card)




                                Vcc:5V, Speed:11Mbps
 PS for 1-hop Ad hoc Networks (1/3)
 Time axis is divided into equal-length intervals
  called beacon intervals
 In the beginning of a beacon interval, there is ATIM
  window, in which hosts should wake up and
  contend to send a beacon frame with the backoff
  mechanism for synchronizing clocks
                    Beacon Interval   Beacon Interval   Beacon Interval   Beacon Interval

                ATIM        Power
               Window       Saving
                            Mode


    Host

           Beacon
PS for 1-hop Ad hoc Networks (2/3)
A   possible sender also sends ATIM (Ad hoc
  Traffic Indication Map) message with DCF
  procedure in the ATIM window to its
  intended receivers in the PS mode
 ATIM demands an ACK. And the pair of
  hosts receiving ATIM and ATIM-ACK should
  keep themselves awake for transmitting and
  receiving data
PS for 1-hop Ad hoc Networks (3/3)
                            Target Beacon Transmission Time (TBTT)

                     Beacon Interval                Beacon Interval

                  ATIM           power         ATIM             active state
                 Window          saving       Window
                                  mode
  Host A
                                             ATIM
      Beacon     No ATIM means                          data
  BTA=2, BTB=5
                 no data to send                       frame
                  or to receive
                  ATIM           power         ATIM
                 Window          saving       Window
                                  mode
  Host B
                                          Beacon ACK           ACK
   PS: m-hop Ad hoc Network
 Problems:
  Clock Synchronization
   it is hard due to communication
   delays and mobility
  Network Partition
   unsynchronized hosts with different
   wakeup times may not recognize
   each other
             Clock Drift Example




Max. clock drift for IEEE 802.11 TSF (200 DSSS nodes, 11Mbps, aBP=0.1s)
Network-Partitioning Example
           C                   D
                   ╳
                                   F    The red ones do not
                                       The blue ones do not
  A            Network                 know the existence of
               Partition
                                         the red ones, not to
                                        the blueones, not to
                   ╳
                                       mention the time when
           B                   E           they are awake.
  Host A
                                                      ATIM
  Host B                   ╳                          window

  Host C

  Host D
               ╳
  Host E

  Host F
Asynchronous PS Protocols (1/2)
Try to solve the network
 partitioning problem to achieve
 Neighbor discovery
 Wakeup prediction
 without synchronizing hosts’ clocks
Asynchronous PS Protocols (2/2)
Three   asyn. PS protocols by Tseng:
 Dominating-Awake-Interval
 Periodical-Fully-Awake-Interval
 Quorum-Based
 Ref:
  “Power-Saving Protocols for IEEE 802.11-Based
  Multi-Hop Ad Hoc Networks,”
  Yu-Chee Tseng, Chih-Shun Hsu and Ten-Yueng Hsieh
  InfoCom’2002
               Outline
 IEEE 802.11 Ad hoc Network
 Power Saving Problem
 Asynchronous Quorum-based PS Protocols
 Optimal Asyn. Quorum-Based PS Protocols
 Analysis and Simulation
 Conclusion
         Numbering beacon intervals
0   1    2    3   4    5    6   7    8   9   10 11 12 13 14 15 0   2 …
              Beacon interval
n consecutive beacon intervals are numbered as 0 to n-1


    0    1        2    3
    4    5        6    7            And they are organized
    8    9        10   11           as a n  n array
    12   13       14   15
         Quorum Intervals (1/4)
Intervals from one row and one column are called

quorum intervals

                          0    1    2    3
Example:
Quorum intervals are      4    5    6    7
numbered by               8    9    10   11
2, 6, 8, 9, 10, 11, 14    12   13   14   15
         Quorum Intervals (2/4)
Intervals from one row and one column are called

quorum intervals

                          0    1    2    3
Example:
Quorum intervals are      4    5    6    7
numbered by               8    9    10   11
0, 1, 2, 3, 5, 9, 13      12   13   14   15
        Quorum Intervals (3/4)
 Any two sets of quorum intervals have two common
 members
For example:
The set of quorum intervals
{0, 1, 2, 3, 5, 9, 13} and        0   1   2  3
the set of quorum intervals       4   5   6  7
{2, 6, 8, 9, 10, 11, 14} have     8   9   10 11
two common members:
                               12   13   14   15
  2 and 9
             Quorum Intervals (4/4)
Host
 D 0     1   2    3    4   5    6    7   8    9   10 11 12 13 14 15
Host
 C 0     1   2    3    4   5    6    7   8    9   10 11 12 13 14 15

                 2 overlapping quorum intervals
Host
 D 0     1   2    3    4    5   6    7   8    9   10 11 12 13 14 15
Host
 C 0      1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
       Even when the beacon interval numbers are not aligned
       (they are rotated), there are always at least two
       overlapping quorum intervals
Structure of quorum intervals
Networks Merge Properly
         C   D

                 F
A


         B   E

Host A
                     ATIM
Host B               window

Host C               Beacon window
                     Monitor window
Host D

Host E

Host F
          Short Summary
There is an asynchronous power-
 saving protocol that achieves
 asynchronous neighbor discovery
  Hearing beacons twice or more in
   every n consecutive beacon intervals
 wakeup prediction
 via a simple quorum concept.
                  Observation 1
 It is a simple grid quorum system [Maekawa 1985]
  in Tseng’s work.
 There are many more complicated quorum
  systems in the literature of distributed system:
     FPP [Maekawa 1985], Tree [Agrawal 1990],
      Hierarchical[Kumar 1991], Cohorts [Jiang 1997], Cyclic
      [Luk 1997], Torus [Lang 1998], etc.
   Question: Can these quorum systems be directly
    applied to solve the power-saving problem in a
    MANET?
            The Answer Is …
 Not   all quorum systems can be used here!
   Counter example: { {1}} under {1,2,3}




 Onlythose quorum systems with the rotation
 closure property can be used!
              Observation 2
 Smaller quorums are better because they imply
  lower active ratio (better energy-efficiency)
 But quorums cannot be too small less the quorum
  system does not satisfy the rotation closure
  property
 Question 1: What is the smallest quorum size?
 Question 2: Is there any quorum systems to have
  the smallest quorum size?
               Outline
 IEEE 802.11 Ad hoc Network
 Power Saving Problem
 Asynchronous Quorum-based PS Protocols
 Optimal Asyn. Quorum-Based PS Protocols
 Analysis and Simulation
 Conclusion
 What are quorum systems?
 Quorum   system:
 a collection of mutually intersecting
 subsets of a universal set U, where
 each subset is called a quorum
  E.G. {{1, 2},{2, 3},{1,3}} is a quorum system
     under U={1,2,3}
A quorum system is a collection of sets
 satisfying the intersection property
 Rotation Closure Property (1/3)
Definition. Given a non-negative integer i
and a quorum H in a quorum system Q under
U = {0,…, n1}, we define rotate(H, i) =
{j+ijH} (mod n).

E.G. Let H={0,3} be a subset of U={0,…,3}.
We have rotate(H, 0)={0, 3}, rotate(H, 1)={1,0},
rotate(H, 2)={2, 1}, rotate(H, 3)={3, 2}
 Rotation Closure Property (2/3)

Definition. A quorum system Q under U =
{0,…, n1} is said to have the rotation closure
property if
G,H  Q, i  {0,…, n1}: G  rotate(H, i)  .
  Rotation Closure Property (3/3)
For   example,
 Q1={{0,1},{0,2},{1,2}} under U={0,1,2} 
 Q2={{0,1},{0,2},{0,3},{1,2,3}} under
  U={0,1,2,3} 
  Because {0,1}  rotate({0,3},3) =
  {0,1}  {3, 2} =        Closure
  Examples of quorum systems
 Majorityquorum system 
 Tree quorum system 
 Hierarchical quorum system 
 Cohorts quorum system 
 ………
   Optimal Quorum System (1/2)
 Quorum  Size Lower Bound for quorum
 systems satisfying the rotation closure
 property:
 k, where k(k-1)+1=n, the cardinality of the
 universal set, and k-1 is a prime power
 (k n )
  Optimal Quorum System (2/2)
 Optimal   quorum system
   FPP quorum system


 Near   optimal quorum systems
   Grid quorum system
   Torus quorum system
   Cyclic (difference set) quorum system
               Outline
 IEEE 802.11 Ad hoc Network
 Power Saving Problem
 Asynchronous Quorum-based PS Protocols
 Optimal Asyn. Quorum-Based PS Protocols
 Analysis and Simulation
 Conclusion
             Analysis (1/3)
 Active Ratio:
  the number of quorum intervals over n,
  where n is cardinality of the universal set
 Neighbor Sensibility (NS)
  the worst-case delay for a PS host to detect
  the existence of a newly approaching PS
  host in its neighborhood
Analysis (2/3)
Analysis (3/3)



                 Optimal!
           Simulation Model
 Area:  1000m x 1000m
 Speed: 2Mbps
 Radio radius: 250m
 Battery energy: 100J.
 Traffic load: Poisson Dist. , 1~4 routes/s,
  each having ten 1k packets
 Mobility: way-point model (pause time: 20s)
 Routing protocol: AODV
               Simulation Parameters
Unicast send            454+1.9 * L
Broadcast send          266+1.9 * L
Unicast receive         356+0.5 * L
Broadcast receive       56+0.5 * L
Idle                    843
Doze                    27                 L: packet length
Unicast packet size           1024 bytes
Broadcast packet size         32 bytes
Beacon window size            4ms
MTIM window size              16ms
       Simulation Metrics
        ratio
Survival
Neighbor discovery time
Throughput
Aggregate throughput
    Simulation Results (1/10)
     E-torus quorum system
                          Cyclic quorum system




Always Active

                    Survival ratio vs. mobility
(beacon interval = 100 ms, 100 hosts, traffic load = 1 route/sec).
                                  Simulation Results (2/10)
Neighbor discovery time (ms)


                                3000

                                2500
                                                                            A faster host can be
                                2000
                                                                               discovered in
                                1500                                            shorter time.
                                1000
                                                C(98)
                                 500
                                                E(7x14)
                                   0
                                            0           5         10          15         20
                                                                pe d   e
                                                        Moving s e (m/s c)
                                            Neighbor discovery time vs. mobility
                               (beacon interval =100 ms, 100 hosts, traffic load = 1 route/sec).
    Simulation Results (3/10)
                   For the throughput: AA>E(7x74)>C(98)




                                       For the aggregate throughput:
                                             C(98)>E(7x74)>AA
                    Throughput vs. mobility
(beacon interval = 100 ms, 100 hosts, traffic load = 1 route/sec).
       Simulation Results (4/10)




                 Survival ratio vs. beacon interval length
(100 hosts, traffic load = 1 route/sec, moving speed = 0~20 m/sec with
                            mean = 10m/sec).
       Simulation Results (5/10)
                                      16000
       Neighbor discovery time (ms)

                                      14000
                                                    C(98)
                                      12000
                                                    E(7x14)
                                      10000
                                      8000
                                      6000
                                      4000
                                      2000
                                         0
                                              100             200          300     400

                                                            Beacon interval (ms)



         Neighbor discovery time vs. beacon interval length
(100 hosts, traffic load = 1 route/sec, moving speed = 0~20 m/sec with
                            mean = 10m/sec).
       Simulation Results (6/10)




                  Throughput vs. beacon interval length
(100 hosts, traffic load = 1 route/sec, moving speed = 0~20 m/sec with
                             mean =10m/sec).
        Simulation Results (7/10)




                       Survival ratio vs. traffic load
(beacon interval = 100 ms, 100 hosts, mobility = 0~20 m/sec with mean =
                                10 m/sec).
        Simulation Results (8/10)




                      Throughput vs. traffic load
(beacon interval =100 ms, 100 hosts, mobility = 0~20 m/sec with mean =
                              10 m/sec).
        Simulation Results (9/10)




                     Survival ratio vs. host density
(beacon interval = 100ms, traffic load 1 route/sec, mobility = 0~20 m/sec
                        with mean= 10 m/sec).
      Simulation Results (10/10)




                      Throughput vs. host density
(beacon interval = 100ms, traffic load 1 route/sec, mobility = 0~20m/sec
                        with mean= 10 m/sec).
               Outline
 IEEE 802.11 Ad hoc Network
 Power Saving Problem
 Asynchronous Quorum-based PS Protocols
 Optimal Asyn. Quorum-Based PS Protocols
 Analysis and Simulation
 Conclusion
               Conclusion
 Quorum    systems with the rotation closure
  property can be translated to an asyn. PS
  protocol.
 The active ratio is bounded by 1/ n, where
  n is the number of a group of consecutive
  beacon intervals.
 Optimal, near optimal and adaptive AQPS
  protocols save a lot of energy w/o degrading
  performance significantly
            Publication

ICPP’03   Best Paper Award

ACM  Journal on Mobile Networks
 and Applications
               Future work
 To incorporate the clustering concept into the
  design of hybrid (syn. and asyn.) power saving
  protocols (NSC 93-2213-E-008-046-)
 To design more flexible adaptive asyn. power
  saving protocols with the aid of the expectation
  quorum system (a novel quorum system which is a
  general form of probabilistic quorum systems)
  (93CAISER-中央大學分部計畫)
 To incorporate power saving mode management
  to wireless sensor networks with comm. and
  sensing coverage in mind (中大新進教師學術研究
  經費補助計畫)
Thanks!
           FPP quorum system
 Proposed  by Maekawa in 1985
 For solving distributed mutual exclusion
 Constructed with a hypergraph
   An edge can connect more than 2 vertices
 FPP:Finite   Projective Plane
   A hypergraph with each pair of edges having
    exactly one common vertex
 Also   a Singer difference set quorum system
    FPP quorum system Example
        5               A FPP quorum
                        system:
                        { {0,1,2},
    3           4         {1,5,6},
            6             {2,3,6},
                          {0,4,6},
0       1           2     {1,3,4},
                          {2,4,5},
                          {0,3,5} }
         Torus quorum system
0    1    2       3   4   5
                               { {1,7,13,8,3,10},
6    7    8       9   10 11    {5,11,17,12,1,14},…}
12 13 14 15 16 17

         One half column cover in a wrap around manner
One full column
For a tw torus, a quorum contains all elements
from some column c, plus w/2 elements, each
of which comes from column c+i, i=1.. w/2
  Cyclic (difference set) quorum system
 Def:  A subset D={d1,…,dk} of Zn is called a
  difference set if for every e0 (mod n), there
  exist elements di and djD such that di-dj=e.
 {0,1,2,4} is a difference set under Z8
 { {0, 1, 2, 4}, {1, 2, 3, 5}, {2, 3, 4, 6}, {3, 4, 5, 7},
  {4, 5, 6, 0}, {5, 6, 7, 1}, {6, 7, 0, 2}, {7, 0, 1, 3} }
  is a cyclic (difference set) quorum system C(8)
  E-Torus quorum system
         Trunk    E(t x w, k)

Branch

                    Branch
                                cyclic
Branch


                 Branch
                                cyclic

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:6
posted:11/1/2011
language:English
pages:74