725 31-10-08 haley jones slides

Upload: ankit-kushwah

Post on 04-Apr-2018

220 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    1/36

    Channel Aware Routing inMANETs with Route Handoff

    Sandra Chen, Haley Jones and Dhammika Jayalath

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    2/36

    Ad-hoc On-demand Distance

    Vector (AODV) Routing Protocol

    On-demand/reactive route discovery

    Single path identified

    RREQ contains:Source IP address (constant)

    Source sequence number (freshness indicator)Broadcast ID (unique)

    Destination IP address (constant)

    Destination sequence number (last known)

    Hop-count (incremented at each intermediate node)

    RREQ uniquely identified by Source SequenceNumber and broadcast ID.

    Duplicate RREQs discarded.

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    3/36

    AODV cont.

    RREP generated when intermediate node has recent

    enough path to destination, or from destination node, itself.

    RREP contains:

    Source IP address (constant)

    Destination IP address (constant)

    Destination sequence number (last known)

    Hop-count (incremented at each node)

    Route expiration time (time for new route discovery)

    Duplicate RREPs discarded, except if same DestinationSequence Number but smaller Hop-Count.

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    4/36

    AODV cont.

    RERR packet generated by up-link node when linkin active route breaks.

    RERR has:

    Source IP address (constant)

    Destination IP address (constant)

    Destination sequence number (incremented from

    RREP)

    Hop-count =

    New route discovery process required by source ifit has remaining packets for destination

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    5/36

    Ad-hoc On-demand Multipath Distance

    Vector (AOMDV) Routing Protocol

    AODV: Destination IP address

    Destination sequencenumber

    Next hop IP address

    Hop-count

    Entry expiration time

    AOMDV: Destination IP address

    Destination sequence number

    Advertised hop-count

    Route list:

    Next hop IP 1, hop-count 1

    Next hop IP 2, hop-count 2

    Entry expiration time

    Similar to AODV, but with Multiple paths identified on each

    route discovery.

    Paths are loop-free and link-disjoint. Routing Table Entries:

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    6/36

    Multiple Link-Disjoint Paths

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    7/36

    Shortcomings of AOMDV

    Paths chosen based only on hop-countPath stability and channel quality measures are

    ignored

    Tends towards paths with few, but long, hopsclose to breaking point

    Channel fading conditions ignored and,therefore, unable to be exploited

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    8/36

    Mobile-to-Mobile Channel Model

    Required because both nodes in a link may be

    moving

    Simple: Uses individual node speeds rather than

    relative speeds

    Has a Rayleigh envelope

    Can define average fading duration (AFD) and

    average non-fading duration (ANFD)

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    9/36

    Fading

    3 2 1 0 1 2 30

    1

    2

    3

    4

    5

    6

    Threshold

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    10/36

    AFD & ANFD

    = ratio of signal threshold to RMS power

    = maximum Doppler shift of transmitter

    = ratio of receiver velocity to transmitter velocity

    ANFD = 1

    fT

    q2(1 + 2)

    AFD =e2 1

    fTq2(1 + 2)

    =Rth

    Rrms

    fT

    =

    vR

    vT

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    11/36

    ANFD as a Link Selection Metric

    Decreases withincreasing normalisedthreshold (indication of

    channel quality andnode separationdistance)

    Decreases with

    increasing (indicationofnode mobility)

    ANFD is a goodindicator oflink

    reliability use asMETRIC

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    12/36

    LMMSE Channel Prediction

    is the predicted signal at time n+kbasedonMincoming signal sample values up until timeinterval n

    is the prediction weight for the ith previous inputsignal sample value

    is the incoming signal sample value at timeinterval n-i

    Precalculate and tabulate w(i) indexed by Doppler shiftand discrete time shift

    x(n + k) = MX

    i=1

    w(i)x(n i)

    x(n + k)

    w(i)

    x(n

    i)

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    13/36

    Channel-Aware AOMDV Protocol

    Use ANFD as measure of path lifetime/duration

    Path duration is defined as the minimum ANFD overall hops in the path

    Implemented by including node speed and directionin RREQ header

    D , min1hH ANFDh

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    14/36

    CA-AOMDV cont.

    AOMDV: Destination IP address

    Destination sequence numberAdvertised hop-count

    Route list:

    Next hop IP 1, hop-count 1

    Next hop IP 2, hop-count 2

    Entry expiration time

    CA-AOMDV: Destination IP address

    Destination sequence numberAdvertised hop-count

    Dmin Route list:

    Next hop IP 1, hop-count 1, D1Next hop IP 2, hop-count 2, D2

    Entry expiration time

    Handoff dormant time

    Comparison of Routing Table Entries:

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    15/36

    CA-AOMDV cont.

    RREPs processed on basis of Destination

    sequence number, Advertised hop-count and Dmin

    Handoff dormant time

    Amount of time a path should be marked as unusableafter handoff

    Set equal to maximum AFD over all links in path

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    16/36

    Path Handoff

    Triggered by prediction of a forthcoming fade

    on one of the path links

    The node at the receiving end of the fading

    link initiates a handoff request via a HREQpacket

    HREQ includes: AFD, Source IP address,Destination IP address and Source sequence

    number

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    17/36

    Path Handoff cont.

    Avoidance of repeat HREQ via maintenance of ahandoff table at each node, similar to routingtable Includes expiration time field to indicate when path is

    expected to again be available for use (out of fade)

    A node forwards a valid HREQ unless it has analternative path to the destination

    When a path has come out of its fade, it may beused again, saving on route discovery overhead,

    unless it has reached its expiry time

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    18/36

    CA-AOMDV Summary

    Route Discovery

    ANFD and hop-count used to select stable, butreasonable length paths

    AOMDV chooses on hop-count only

    Route Maintenance

    Predicted signal strength used for handoff

    initiation

    AFD used to choose when to bring path out of

    dormancy

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    19/36

    Theory

    Average number of hops between 2nodes, with transmission rangeR:

    Average number of hops before

    encountering broken link is

    With Cconnections at any time, andn neighbours, average number ofconnections over a given link is

    s

    s

    N nodes in network

    H =S(

    2 + l n ( 1 +

    2))

    3R.

    (H + 1)/2

    B = 2CH

    nN.

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    20/36

    Single Link and Path Lifetimes

    Link lifetime,Zl, has exponential distribution

    Assume a path withL i.i.d. links. Path lifetime,Zp,also has exponential distribution

    Expected Path Lifetime

    fZ

    (t) = `e`t

    fZp(t) = et

    PLk=1 k

    L

    X`=1

    ` = pept = LeLt

    E{Zp} =Z

    0tpeptdt = 1

    p= 1

    L

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    21/36

    AOMDV Multiple Path System Lifetime

    Assume all paths haveL i.i.d. links.

    The AOMDV multiple path system lifetime isZA.

    System is up as long as any of theNp paths are still up(though once down they are discarded)

    Expected lifetime of AOMDV multiple path system ANFD

    Pr{ZA < t} = Pr{(Zp1 < t)

    (Zp2 < t)

    (ZpNp < t)}

    = (1 ept)Np = FZA(t).

    E{ZA} = Npp

    Xk=0

    (1) Np

    1

    Np k 1

    Np 1k

    !1

    (Np k)2.

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    22/36

    Single Link and Path Downtimes

    Link downtime, Yl, has exponential distribution

    Assume a path withL i.i.d. links. Path downtime, Yp,also has exponential distribution.

    The CA-AOMDV multiple path system downtime is

    YC

    FY (t) = 1 et

    CA AOMDV Multiple Path System

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    23/36

    CA-AOMDV Multiple Path System

    Downtime

    The CA-AOMDV system is down only when all paths

    are down.

    Expected System Downtime AFD

    Pr{YC > t} =

    NpYi=1

    Pr{Ypi > t} =

    NpYi=1

    1

    LYk=1

    FYk

    (t)

    =

    1

    1

    etLNp

    FYC(t) = 1 1

    1 et

    L

    Np

    {YC} = NpL

    Np1Xk=0

    (1)Npk1

    Np

    1k

    !L(Npk)1Xi=0

    (1)L(Npk)1i

    L(Np

    k)

    1i

    !1

    (L[Np k] i)2

    .

    CA AOMDV Multiple Path System

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    24/36

    CA-AOMDV Multiple Path System

    Lifetime

    Probability that a path is down is the probability that

    at least one link is in a fade.

    Recall that a link signal strength has Rayleighdistribution, with parameter, .

    AFD)fadeinischannelPr(

    )fadeinischannelPr(1

    RateCrossingLevel)fadeainnotischannelPr(ANFD

    RateCrossingLevel

    )fadeainischannelPr(AFD

    =

    =

    =

    CA-AOMDV Multiple Path

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    25/36

    CA-AOMDV Multiple Path

    System Lifetime cont.

    Based on the Rayleigh distribution of the link gains,

    probability CA-AOMDV system in fade is

    So, ifZC is CA-AOMDV system lifetime

    Pr(system in fade) =

    Np

    Yi=1

    1

    L

    Y`=1

    e2

    = 1 e

    L2

    Np

    E{ZC} =1

    1 eL2

    Np1 eL

    2NpE{YC}

    AOMDV vs CA-AOMDV Multiple

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    26/36

    AOMDV vs CA-AOMDV Multiple

    Path System Lifetimes

    5 4 3 2 1 0 1 20

    0.5

    1

    1.5

    2

    (dB)C

    A

    AOMDVtoA

    OMDVlifetimer

    atio

    Np

    = 2

    L=3L=4

    L=5

    L=6

    5 4 3 2 1 0 1 20

    0.5

    1

    1.5

    2

    (dB)C

    A

    AOMDVtoA

    OMDVlifetimer

    atio

    Np

    = 6

    L=3L=4

    L=5

    L=6

    AOMDV vs CA-AOMDV Multiple

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    27/36

    AOMDV vs CA AOMDV Multiple

    Path System Lifetimes

    5 4 3 2 1 0 1 20

    0.5

    1

    1.5

    2

    (dB)C

    A

    AOMDVtoA

    OMDVlifetimer

    atio

    Np

    = 2

    L=3L=4

    L=5

    L=6

    25 20 15 10 5 0 50

    1

    2

    3

    4

    5

    6

    7

    8

    threshold, (dB)

    CAAOMDVtoA

    OMDVLIfetimeRatio

    Np

    = 2, L = 1

    AOMDV vs CA-AOMDV Multiple

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    28/36

    AOMDV vs CA AOMDV Multiple

    Path System Lifetimes

    25 20 15 10 5 0 50

    50

    100

    150

    200

    250

    300

    350

    (dB)

    MultiplePathSy

    stemL

    ifetime

    Np

    = 2, L = 1

    CAAOMDVAOMDV

    20 15 10 5 0 50

    2

    4

    6

    8

    10

    12

    14

    16

    18

    20

    (dB)

    MultiplePathSystemL

    ifetime

    Np

    = 2, L = 1

    CAAOMDVAOMDV

    R ti C t l O h d

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    29/36

    Routing Control Overhead

    C = connections in the network

    T = average connection time

    = route discoveries per second per node pair N = node degree

    Np = number of saved paths

    Nrq = number of RREQ per path repair B = average number of paths over a breaking link

    = average hop-count

    A = CTA(N + NpH) + CT NpAB

    3H + 2

    4+

    Nrq

    2

    !

    C

    =CT

    C(

    N+

    NpH

    ) +CT Np

    CB

    I

    H

    D l

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    30/36

    Delay

    d = delay due to route discovery from failed system

    tr= one-hop propagation time of RERR

    td = one-hop propagation time of a data packet

    th = one-hop propagation time of HREQ

    TMAC = channel occupation time due to MAC overhead

    = expected number of hops to deliver HREQ

    A = dAT + T N

    pA(tr + TMAC)(H + 1)2

    + T NpANR(td + TMAC) + TATMACH

    C = dCT + T NpC(th + TMAC)I

    + TCNR(td + TMAC) + TCTMACH

    I

    P k t D li R ti

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    31/36

    Packet Delivery Ratio

    q = probability that channel is free for transmission

    ( )

    ( )T

    qTT

    qT

    C

    C

    AA

    =

    =

    Th h t P k t R t

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    32/36

    Throughput vs Packet Rate

    33% improvement at

    higher packet rates

    Choice of stable links in

    CA-AOMDV, plus

    handover fewer route

    discoveries

    Thro ghp t s Mobilit

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    33/36

    Throughput vs Mobility

    ~27% improvement at

    7m/s (25km/h)

    Delay vs Mobility

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    34/36

    Delay vs Mobility

    About 25% improvement

    at v = 2.5 m/s

    Routing Overhead vs Mobility

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    35/36

    Routing Overhead vs Mobility

    18% improvement at

    v = 2.4m/s

    Conclusions

  • 7/29/2019 725 31-10-08 Haley Jones Slides

    36/36

    Conclusions

    Can improve AOMDV performance byTaking into account channel conditions in addition to hop-

    count

    Using channel prediction to initiate path hand-off