crn - assisted multicasting-2011!12!20

Upload: kashif-aziz-awan

Post on 06-Apr-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    1/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN)

    Efficient Multicasting in CognitiveRadio Networks Using Cooperation

    and Channel Diversity

    Ahmed E. Kamal

    Dept. of Electrical & Computer EngineeringIowa State University

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    2/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 22

    Outline

    Cognitive radio networks (CRN): Why?

    What?

    Multicasting problem in CRNs

    Receiver assisted multicasting (RAM) for CR-WMNs: The single cell case

    The multiple cell case (collision free schedules)

    Summary and conclusions

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    3/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 33

    Motivation Behind Opportunistic Spectrum Access Initiative

    The current spectrum allocationpolicy leads to:

    Spectrum crowdedness

    Spectrum underutilization

    Opportunistic spectrum access(OSA)

    Secondary users (SUs) getopportunistic access to idleunlicensed channels.

    Secondary users must not causeharmful interference to licensedusers (Primary Users, or PUs)

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    4/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 44

    Opportunistic Spectrum Access (OSA)

    OSA requires that SUs:

    Monitor spectrum activity

    Switch between frequency

    channels, probably at a fine

    granularity, as needed.

    Software Defined Radios is the enabling

    technology ofOSA.

    Software defined radios are able to change

    their transmission/reception parameters like:

    Operating frequency

    Transmission power

    Modulation scheme

    Cognitive Radio is a software defined radio radio that can change its operationalparameters in response to changes in the surrounding wireless environment.

    [1] Cognitive Radio: An Integrated Agent Architecture for Software Defined Radio. Dissertation.

    Doctor of Technology. Joseph Mitola III. Royal Institute of Technology (KTH), May2000

    .

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    5/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 55

    A Cognitive Radio Network (CRN)

    Is a wireless network that relies onO

    SA for its operation. Research directions in CRN:

    Sensing, and identifying idle spectrum channels.

    Spectrum sharing and management

    Networking protocols specific to CRNs.

    After identifying idle spectrum channels, a CRN is amultichannel wireless networks, with different channelavailabilities at different locations, and different times

    However, there are some major difference between CRNs andtraditional multichannel wireless networks.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    6/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN)

    Assisted Multicast Scheduling

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    7/44

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    8/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 88

    Network Model: Cognitive Radio Wireless Mesh Network (CR-WMN)

    CR operates as a Wireless Mesh Network

    (WMN)

    The CR-WMN consists of multiple meshrouters (MR):

    Each MR manages a groups of mesh clients (MC).

    MR and its clients MC form a cell

    MRs in adjacent cells communicate The CR-WMN coexists with a primary

    network, using one or more licensed channels

    The CR-WMN opportunistically exploits theset of licensed channels, where the availability

    of each of which is determined by a SpectrumSensing and Management Entity (SSME).

    A gateway, or more, exits which connects theCR-WMN to the Internet.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    9/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 99

    The Multicast Problem

    We first consider the multicast scheduling problem within asingle cell of a CR-WMN

    We then extend this to guarantee collision free scheduling acrossmultiple cells.

    Mmulticast sessions exist in the cell, all of which originate from

    the MR.

    The network opportunistically utilizes a set Lof orthogonal

    channels, with K=|L|

    The nodes share a control channel or any of the alternativesproposed in literature for control information exchange.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    10/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 1010

    Multicasting in CRNs: Broadcast Deformation

    Channel heterogeneity problem in CRNs

    Broadcasting using a single transmission may not be possible inCRNs.

    In the example below, SU A needs two transmissions to broadcast

    the same data to the other three SUs.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    11/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 1111

    Proposed Solution

    Cooperative multicasting: MCs which receive packets, assist MR in delivering packet to other

    receivers (i.e., assisted multicast):

    Within the same group (intra-group assistance)

    Between different groups and using overhearing (inter-group assistance)

    Network coding & packet overhearing (between multicast groups)

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    12/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 1212

    Unassisted Multicast

    - ato white nodes

    - bto black nodes

    Non-cooperativemulticast requires sixtime slots in this case.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    13/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 1313

    Assisted Multicast Within The Same Group (Intra-Group)

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    14/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 1414

    Assisted Multicast across Multiple Groups (Inter-Group)

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    15/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 1515

    Network Coding Assisted Multicast (Codeword Exchange)

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    16/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 1616

    Assisted Multicast Scheduling Problem

    The unassisted multicast scheduling problem (UMS) is similar to AMSexcept that no assistance or codeword exchange operations areallowed.

    The UMS and AMS problems are NP-hard (even for a single group): By reduction from set cover problem

    Two ILPs were introduced for the UMS and AMS problems with a

    single multicast group.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    17/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 17

    ILP of the AMS-Single Cell, Single Group Problem

    Input: a single multicast group, channel availability at each member,connectivity information between members.

    Objective: minimize the multicast period.

    Constraints:

    At most one transmission per node per time slot.

    At most one transmission per channel per time slot.

    A node can either receive or transmit in any time slot.

    A node (except the MR) cannot transmit a packet at time t unless it has receivedit in [0,t-1].

    Output:A multicast schedule in the form of:(node, codeword, channel, slot)

    17

    UMS formulation is straight forward (set cover)

    AMS-Multihop is too complicated (mainly because of the coding operating)

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    18/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 1818

    Heuristic Assisted-Multicast Scheduling (HAMS) Algorithm

    HAMS deals, iteratively, with each slot independently andtries to optimize the schedule within that slot.

    Within each slot:

    1. Scheduling the MR transmission (channel and codeword):

    Select the codeword that covers maximum number of uncovered MCs.2. Scheduling the assistance operation (channel and codeword):

    Select the (channel, codeword) pair that covers maximum number ofuncovered neighbors.

    3. Scheduling overhearing opportunities:

    The codeword that is useful to maximum number of uncovered neighborsis overheard.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    19/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN)

    Performance Evaluation

    19

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    20/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 2020

    Gain of Different Assistance Levels

    Gain = 100*(Unassisted Assisted )/Unassisted

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    21/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 21

    HAMS Performance (Single-group)

    21

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    22/44

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    23/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 23

    Effect of Channel Availability

    23

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    24/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN)

    The Multiple Cell Case:Collision Avoidance

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    25/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 2525

    Collision Free Scheduling across Cells

    How to make sure that the schedule in one cell of the mesh network

    has no conflict (collision) with the schedule of any of the adjacentcells?

    Reactive: calculate the schedules and then resolve collisions Pros:

    Each cell can make full use of the available channels, obtaining higher gainfrom the assistance mechanism.

    Cons: Requires a collision resolution mechanism, and difficult to implement.

    Proactive: proactively build collision free schedules. Pros:

    Easier to implement, and requires no post-scheduling collision resolution.

    Cons: A Cell may not be able to utilize the full set of available channels, and will

    thus have a smaller assistance gain.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    26/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 2626

    Collision Free Scheduling across Cells: Reactive Approach

    Calculate the schedules of all cells.Align the schedules such that there are no collisions

    and the time span (time to schedule all cells) is

    minimized (ILP formulation).

    The alignment considers the following properties: Precedence property: an MC cannot transmit a codeword unless

    it has already received it (or a combination that can produce it)

    through an earlier transmission(s). Conflict property:

    One transmission per node per slot.

    One transmission per channel.

    A node can either a transmitter or a receiver in a give time slot.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    27/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 2727

    Collision Free Scheduling across Cells: Proactive Approach

    MRs exchange requests to start scheduling.

    Priority mechanism is used to break ties in case ofconcurrent requests.

    When the MR receives acknowledgments from all of its

    neighboring MRs, it calculates its schedule and sends it to itsneighbors.

    MRs will avoid using a particular channel, in a particularslots, if it is used in a neighboring cell at the same time.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    28/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 28

    Proactive vs. Reactive

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    29/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 29

    Summary and conclusions

    We proposed an assistance paradigm that relies on receivernodes to forward the multicast data to other receivers thathave not yet received their own data.

    Network coding was also proposed as another assistance

    technique that further reduced the total multicast period. Cooperative networking solutions are a must for networks

    with heterogeneous and unguaranteed resources like CRNs.

    Any networking solution with no mechanism to recover

    from channel failures, caused by licensed users activity, isan incomplete solution in the case of CRNs.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    30/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN)

    Acknowledgements

    Dr. Hisham Almasaeid (currently with Amazon)

    Financial support by the National Science Foundation, USA

    30

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    31/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 31

    Thanks..Questions?

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    32/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN)

    Cognitive Radio Networks vs. TraditionalMultichannel Wireless Networks

    32

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    33/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 3333

    Heterogeneous Channel Availability, Intermittent Connectivity

    Different SUs may observe

    different sets of availablechannels depending on theactivity of primary, licensed,users (PUs) in their vicinity.

    Channel availability can changespatially and temporally.

    This is referred to as the channelavailability heterogeneity property.

    Intermittent connectivity, evenfor stationary SUs.

    Harmful

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    34/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 3434

    Dynamic Topology

    In traditional networks,

    network topology is stable forstationary nodes. This may notbe the case in CRNs.

    Rerouting can induce significant

    overhead.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    35/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 3535

    Broadcast Deformation

    Broadcasting using a single transmission may not be possible in

    CRNs.

    In the example below, SU A needs two transmissions to broadcast

    the same data to the other three SUs.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    36/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 3636

    Heterogeneous Transmission Range

    In some proposals, SUs can optimize their transmission powers to

    make the most use of spectrum resources.

    This leads to heterogeneity in the transmission range across different channelsfor the same SU.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    37/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 3737

    Communication Coordination

    How to coordinate the communication and medium access

    between SUs?

    Common Control Channel

    May not always exist

    Subject to congestion

    Source of system vulnerability, attacking this channel can cripple the network.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    38/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 3838

    Switching Latency

    Cognitive radio nodes might have to switch between widely separated

    channels. Channel switching latency becomes a key concern.

    Cannot be ignored while making routing decisions.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    39/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN)

    Failure Recovery

    39

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    40/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 4040

    Recovery From Failures

    There is a need for a recovery mechanism to fix theschedule in case of any interruptions caused by PU activity.

    If an MR transmission fails, it can retry it in the next timeframe or add extra slots to the time frame (on available

    channels) How to recover MC transmissions?

    Assume an MC zis supposed to transmitp1+p2 at slot t, giventhat it has receivedp1 at t-1 from xandp2 at t-2fromy. If either

    of the transmissions scheduled at t-1 and t-2fails, the transmissionat twill be delayed.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    41/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 4141

    Queueing System at MCs

    Input Queue (IQ): holds received codewords, indexed by frame ID.

    Availability Queue (AQ): a virtual queue holding codewords which can bedecoded from the ones in IQ. This queue is indexed by frame ID.

    Delayed Queue (DQ): a virtual queue holding all scheduled codewords which arenot yet in AQ. Indexed by slot ID.

    Output Queue (OQ): holds scheduled codewords which are in AQ, but not yet

    transmitted due to channel unavailability.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    42/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 4242

    Recovery Scheduling at the MR

    MCs inform the MR about the size of their OQs.

    Whenever there are some backlogged MCs, the MRcalculates a recovery schedule to run in the next frame.

    The MR processes the backlogged MCs iteratively, giving

    priority to the one with the maximum OQ size. It assigns recovery slots, if possible, for the selected transmission,

    and moves to the next MC.

    If no recovery slots are found, nothing is done. If the packet delay

    exceeds a predetermined threshold, the MR recalculates the wholeschedule.

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    43/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 4343

    Example: original schedule

  • 8/3/2019 CRN - Assisted Multicasting-2011!12!20

    44/44

    IOWA STATE UNIVERSITY LABORATORY FOR ADVANCED NETWORKS (LAN) 4444

    Example: recovery behavior