Broadcast Strategies Meghanathan 76

download Broadcast Strategies Meghanathan 76

of 12

Transcript of Broadcast Strategies Meghanathan 76

  • 8/7/2019 Broadcast Strategies Meghanathan 76

    1/12

    Performance Studies of MANET Routing Protocols in the Presence of

    Different Broadcast Route Discovery Strategies

    Dr. Natarajan Meghanathan

    Department of Computer Science

    Jackson State UniversityJackson, MS 39217

    Email: [email protected]

    ABSTRACT

    Simulation studies for the Mobile Ad hoc NETwork (MANET) routing protocols have so

    far employed flooding as the default mechanism of route discovery. During flooding,each node broadcasts the packet exactly once, causing the broadcast storm problem [1].

    Several efficient broadcasting strategies [1][2] that reduce the number of retransmitted

    route query packets and the number of retransmitting nodes have been proposed in theliterature. These include the probability-based, area-based and neighbor-knowledge

    based methods to reduce the retransmission overhead. Our contribution in this paper is anns-2 simulation based analysis on the impact of employing these broadcasting strategies

    for route discovery on the hop count and stability of routes. We use the minimum-hop

    based Dynamic Source Routing (DSR) protocol [3] and the stability-based Flow-Oriented Routing Protocol (FORP) [4] as the routing protocols for our analysis. We

    compare the hop count and stability of DSR and FORP routes determined under

    conditions that guarantee at least 92-95% success in route discoveries and simultaneouslyminimize the number of retransmissions and retransmitting nodes.

    Keywords: Broadcasting, Routing, Stability, Hop count, Mobile Ad Hoc

    Networks

    1 INTRODUCTIONA Mobile Ad hoc NETwork (MANET) is a

    dynamic distributed system of autonomously moving

    wireless nodes (such as laptops, personal digitalassistants, etc) and lacks a fixed infrastructure. The

    network has limited bandwidth as the wireless

    medium is shared and is prone to transmissioninterference. Nodes are battery-powered and

    operated with a limited transmission range. As a

    result, routes in MANETs are often multi-hop in

    nature and have to be discovered by the nodesthemselves. There is no centralized administration

    like in cellular networks. Several unicast and

    multicast MANET routing protocols have beenproposed in the literature. The route discovery could

    be either proactive or reactive. In the proactive

    approach, nodes determine and maintain routes for

    every possible source-destination pair, irrespective oftheir requirement. Reactive or on-demand MANETrouting protocols determine a route only when

    required. It has been observed [5][6] that with adynamically changing network topology where route

    accuracy and routing overhead are crucial, on-

    demand routing protocols are to be preferred over the

    proactive protocols. We will focus only on on-

    demand routing for the rest of this paper.Currently, all the on-demand MANET routing

    protocols employ a simple form of broadcasting

    called flooding to discover the routes. Whenever a

    source node has data to send to a destination node,

    but does not have the route to the same, it will

    initiate a broadcast route-query process. In the case

    of flooding, the source node broadcasts a Route-Request-Query (RREQ) packet to its neighbors. Each

    node in the network will broadcast this RREQ packet

    exactly once when they see it the first time. Thedestination node receives the RREQ packets along

    several paths, chooses the best route according to the

    route selection principles of the particular routing

    protocol and notifies the source node about theselected route using a Route-Reply (RREP) packet.

    Flooding is a very expensive process with respect

    to the bandwidth and energy usage. With resource-constrained environments like those of MANETs,

    employing flooding for on-demand route discovery

    will be very costly. Flooding also introduces lot of

    redundancy in the packet retransmission process. In[1], it has been observed that with flooding, when anode receives a packet for the first time, at least 39%

    of the nodes neighborhood would have also receivedthe message simultaneously and on average only

    41% of additional area could be covered with a

    rebroadcast. In general, when a node rebroadcasts a

    message after hearing it k times, the expected

    additional coverage decreases exponentially withincreasing values of k [1]. These observations

    motivated researchers to introduce several efficient

  • 8/7/2019 Broadcast Strategies Meghanathan 76

    2/12

    broadcasting strategies that will minimize thenumber of redundant retransmissions and at the same

    time maximize the chances of the broadcasted

    message reaching all the nodes in the network.The techniques for efficient broadcasting can be

    grouped into three families [1][2]: probability-based

    methods, area-based methods and the neighbor

    knowledge-based methods. In probability-basedmethods, each node is assigned a probability for

    retransmission. In area-based methods, a common

    transmission range is assumed and a node willrebroadcast if only sufficient new area can be

    covered with the retransmission. In neighbor-

    knowledge based methods, each node stores

    neighborhood state information and uses it to decidewhether to retransmit or not. One or more

    broadcasting techniques have been proposed under

    each of the above three families. The objective of allthese broadcasting techniques is to minimize the

    number of retransmitted messages and the number of

    nodes retransmitting the message. More information

    on the different broadcasting techniques can be

    found in Section 3.The performance of the different efficient

    broadcasting techniques under different conditions of

    topology changes and offered broadcast traffic hasbeen studied in [2]. As the number of retransmitting

    nodes and the retransmitted messages get reduced

    when using these broadcasting techniques for RREQ

    propagation, the quality of the routes chosen may bedifferent compared to those routes discovered using

    simple flooding. This formed the motivation for us toimplement these broadcasting techniques and use

    them for route discovery in on-demand MANET

    routing protocols.

    On-demand MANET routing protocols can be

    classified into two broad categories [7]: minimum-weight based routing protocols and stability-oriented

    routing protocols. The Dynamic Source Routing

    (DSR) protocol [3] is a well-known minimum-weight based protocol that selects routes with the

    minimum hop count. The Flow-oriented Routing

    Protocol (FORP) [4] was observed to discover the

    most stable routes within the class of stable pathrouting protocols [8]. The stability of routes selected

    by a routing protocol is quantified in terms of the

    number of route transitions incurred by the protocolfor a source-destination (s-d) session. More

    information on DSR and FORP is provided in

    Section 2.

    In this paper, we implement the probability- based method, the distance-based technique (area-based method), the Multi-Point Relaying (MPR) and

    the Minimum Connected Dominating Set (MCDS)

    based techniques (neighbor-knowledge basedmethod) as the route discovery strategies for DSR

    and FORP and study the impact of these

    broadcasting techniques on the quality of routeschosen by the two routing protocols. We specifically

    study the impact on two principal routing metrics,viz., the stability and hop count. We compare the

    stability and hop count of DSR and FORP routes

    chosen with these broadcasting techniques with thosediscovered using flooding. Flooding helps to

    discover the minimum hop routes for DSR and the

    most stable routes for FORP. But, these efficient

    broadcasting techniques may not yield the minimumhop routes for DSR or the most stable routes for

    FORP.

    The rest of the paper is organized as follows: InSection 2, we briefly discuss the DSR and FORP

    protocols. Section 3 discusses the different

    broadcasting techniques that have been published in

    the literature. Section 4 describes the simulationenvironment, illustrates the results and interprets

    them. Section 5 concludes the paper. Note that we

    use the words route and path, message andpacket, rebroadcast and retransmit

    interchangeably in this paper.

    2 REVIEW OF MANET ROUTINGPROTOCOLS

    In this section, we briefly review the minimum-

    hop based Dynamic Source Routing (DSR) protocol[3] and the stability-based Flow-Oriented Routing

    Protocol (FORP) [4] the two protocols we use for

    our simulation analysis.

    2.1 Dynamic Source Routing (DSR) ProtocolThe unique feature of DSR [3] is source routing:

    data packets carry the route from the source to the

    destination in the packet header. As a result,

    intermediate nodes do not need to store up-to-date

    routing information. This avoids the need for beaconcontrol neighbor detection packets that are used in

    the stability-oriented routing protocols. Route

    discovery is by means of the broadcast query-replycycle. A source nodes wishing to send a data packet

    to a destination d, broadcasts a Route-Request

    (RREQ) packet throughout the network. The RREQ

    packet reaching a node contains the list ofintermediate nodes through which it has propagated

    from the source node. After receiving the first RREQ

    packet, the destination node waits for a short timeperiod for any more RREQ packets and then chooses

    a path with the minimum hop count and sends a

    Route-Reply Packet (RREP) along the selected path.

    If any RREQ is received along a path whose hopcount is lower than the one on which the RREP wassent, another RREP would be sent on the latest

    minimum hop path discovered.

    2.2 Flow-Oriented Routing ProtocolFORP [4] utilizes the mobility and location

    information of nodes to approximately predict the

  • 8/7/2019 Broadcast Strategies Meghanathan 76

    3/12

    Link Expiration Time (LET) for each wireless link.FORP selects the route with the maximum Route

    Expiration Time (RET), which is the minimum of the

    LET values of the constituent links of the route. Eachnode periodically sends a beacon control message to

    its neighbors and the message includes the current

    position of the nodes, velocity, the direction of

    movement and the transmission ranges. Each node isassumed to be able to predict the LET values of each

    of its links with the neighboring nodes based on the

    information collected using beacon packets. FORPassumes the availability of location identifying

    techniques like GPS (Global Positioning System) [9]

    and also assumes that the clocks across all nodes are

    synchronized.Given the motion parameters of two neighboring

    nodes, the duration of time the two nodes will remain

    neighbors can be predicted as follows: Let two nodesi andjbe within the transmission range of each other.

    Let (xi, yi) and (xj, yj) be the co-ordinates of the

    mobile hosts i and j respectively. Let vi, vj be the

    velocities and i, j, where (0 i, j< 2) indicate

    the direction of motion of nodes i andj respectively.The amount of time the two nodes i and j will stay

    connected,Di-j, can be predicted as follows:

    Dab cd a c r ad bc

    a ci j =

    + + +

    +

    ( ) ( ) ( )2 2 2 2

    2 2

    where,a = vi cosi vj cosj; b = xi xj; c = vi sini vjsinj; d = yi yj

    Route discovery is accomplished using the broadcast query-reply cycle with RREQ packets

    propagating from the source node s to the destination

    node don several paths. The information recorded inthis case by a node i receiving a RREQ messagefrom a nodej is the predicted LET of the linki-j. The

    destination d will receive several RREQ messages

    with the predicted LETs in the paths traversed beinglisted. Thes-dpath that has the maximum predicted

    RET is then selected. If more than one path has the

    same maximum predicted RET, the tie is broken by

    selecting the minimum hop path of such paths.

    3 REVIEW OF BRAODCASTINGSTRATEGIES

    In general, the broadcasting strategies can be

    grouped into four families: Simple flooding,Probability-based methods, Area-based methods and

    Neighbor knowledge based methods.

    3.1 Simple FloodingA source node initiates flooding by broadcasting

    a packet to all its neighbors. The neighbor nodes in

    turn rebroadcast the packet exactly once and the

    process continues until each node in the network hasretransmitted the packet. As a result, all nodes

    reachable from the source receive the packet.

    Flooding causes the broadcast storm problem [1]which is characterized by redundant rebroadcasts,

    channel contention and collision of messages.

    3.2

    Probability-based Methods

    3.2.1 Probabilistic SchemeWhen a node receives a broadcast message for

    the first time, the node rebroadcasts the message with

    a probability P. If the message received is already

    seen, then the node drops the message irrespective of

    whether or not the node retransmitted the messagewhen received for the first time. For sparse networks,

    the value ofPhas to be high enough to facilitate a

    higher packet delivery ratio. WhenP= 1, the schemeresorts to simple flooding.

    3.2.2 Counter-based SchemeA broadcast message received for the first time is

    not immediately retransmitted to the neighborhood.The message is queued up for a time called the

    Random Assessment Delay (RAD) during which the

    node may receive the same message (redundantbroadcasts) from some of its other neighbors. After

    the RAD timer expires, if the number of times the

    same message is received exceeds a counter

    threshold, the message is not retransmitted and issimply dropped.

    3.3 Area-based Methods3.3.1 Distance-based Scheme

    When a node receives a previously unseen

    broadcast message, the node computes the distancebetween itself and the sender. If the sender is closer

    than a threshold distance, the message is dropped and

    all future receptions of the same message are alsodropped. Otherwise, the received message is cached

    and the node initiates a RAD timer. Redundant

    broadcast messages received before the expiry of the

    RAD timer are also cached. When the RAD timerexpires, the node computes the distance between

    itself and the neighbor nodes that previously

    broadcast the particular message. If any suchneighbor node is closer than a threshold distance

    value, the message is dropped. Otherwise, the

    message is retransmitted.

    3.3.2 Location-based SchemeWhenever a node originates or rebroadcasts a

    message, the node puts its location information in the

    message header. The receiver node calculates theadditional coverage area that would be obtainable if

    it were to rebroadcast. If the additional coverage is

    less than a threshold value, all future receptions ofthe same message will be dropped. Otherwise, the

  • 8/7/2019 Broadcast Strategies Meghanathan 76

    4/12

    RAD timer is started. Redundant broadcast messagesreceived before the expiry of the RAD timer are also

    cached. After the RAD timer expires, the node

    considers all the cached messages and recalculatesthe additional obtainable coverage area if it were to

    rebroadcast the particular message. If the additional

    obtainable coverage area is less than a threshold

    value, the cached messages are dropped. Otherwisethe message is rebroadcast.

    3.4 Neighbor Knowledge based Methods3.4.1 Multi-point Relaying

    Under this scheme, each node is assumed to have

    a list of its 1-hop and 2-hop neighbors, obtained via periodic Hello beacons. The Hello messages

    include the identifier of the sending node, the list of

    the nodes known neighbors and the Multi-PointRelays (MPRs). After receiving Hello messages

    from all its neighbors, a node has the 2-hop topology

    information centered at itself. Using this list of 1-hop

    and 2-hop neighbors, a node selects the MPRs the

    1-hop neighbors that most efficiently reach all nodeswithin its 2-hop neighborhood. Each node selects the

    set of MPRs using a greedy approach of iteratively

    including the 1-hop neighbors that would cover thelargest number of uncovered 2-hop neighbors.

    3.4.2 Minimum Connected Dominating SetA Connected Dominating Set (CDS) is a set of

    nodes in the network such that all nodes in the

    network are either in the CDS or directly attached toa node in the CDS. A Minimum Connected

    Dominating Set (MCDS) is the smallest CDS, in

    terms of the number of nodes in the CDS, for the

    entire network. The size of the MCDS is the

    minimum number of retransmissions required in abroadcasting process so that all nodes in the network

    receive the broadcast message. Determining the

    MCDS for a given network graph is an NP-complete problem and hence several heuristics have been

    proposed to approximate the MCDS for a given

    network graph.

    4 SIMULATIONSWe use ns-2 (version 2.28) [10] as the simulator

    for our study. The network dimensions are 1500m x

    300m. The transmission range of each node is 250m.

    These values are very commonly used in MANET

    simulations. We vary the density of the network byconducting simulations with 25 nodes (low density)and 50 nodes (high density). The simulation time is

    1000 seconds. While we implemented the FORP

    protocol, we used the implementation of DSR thatcomes with ns-2.

    4.1 Broadcasting Strategies

    The route discovery mechanism in each of DSRand FORP is implemented with the following

    broadcasting strategies: Simple flooding,

    Probabilistic broadcasting, Distance-basedbroadcasting, MPR and MCDS-based broadcasting.

    We choose the probabilistic scheme over counter-

    based scheme as the range of counter values to

    experiment with changes dynamically depending onnetwork density and node mobility. We choose the

    distance-based scheme over the location-based

    scheme because of the higher overhead in computingthe additional coverage area when a node receives

    multiple broadcast messages from its neighbors. The

    probability of retransmission was varied from 0.1 to

    1. The threshold distance for triggering a broadcast isvaried from 20m to 200m, in increments of 20m. We

    do not let any intermediate node to reply for the

    RREQ packets and disable local route repairs as thismay affect our goal on studying the effect of the

    different broadcasting strategies on the routing

    metrics. We do not expect much congestion in our

    network scenarios. Hence, the value of the RAD

    timers used for the distance-based scheme is 0.01seconds, as suggested in [2].

    4.2 Beacon MessagingEach node periodically broadcasts a Hello

    beacon message in its neighborhood. The Hello

    message contains the following information: thelocation of the node, its velocity and direction of

    moving, the 1-hop neighbor list of the node, and theset of MPRs for the node. The Hello message is

    used by FORP and the MPR and MCDS based

    broadcasting strategies. In the case of FORP, the

    clocks across all nodes are assumed to be

    synchronized and each node also keeps track of thepreviously advertised location of its neighbor nodes.

    This helps to keep track of the direction in which the

    neighbor node is moving.

    4.3 Simulation ModelsThe physical, data link and MAC layer models

    are based on the multi-hop wireless network

    extension [5] provided by the CMUs Monarch

    research group. The MAC layer uses the DistributedCoordinated Function (DCF) of the IEEE Standard

    802.11 [11] for wireless LANs. The radio model uses

    the standard channel bandwidth of 2Mbps. The

    signal propagation model used is the two-ray groundreflection model [5]. The interface queue stores boththe routing and data packets sent by the routing layer

    until the MAC layer is able to transmit them. We use

    a FIFO-based interface queue of length 100.The node mobility model used is the Random

    Waypoint model [12]. Each node starts moving from

    an arbitrary location to a randomly selecteddestination location at a speed uniformly distributed

  • 8/7/2019 Broadcast Strategies Meghanathan 76

    5/12

    in the range [vmin, , vmax]. Once the destination isreached, the node may stop there for a certain time

    called the pause time and then continue to move by

    choosing a different target location and a differentvelocity. In this paper, we set vmin = 0. The vmax

    values are 5, 10 (low mobility), 20, 30 (moderate

    mobility), 40 and 50m/s (high mobility). The pause

    time is 0 seconds.

    4.4 Performance MetricsWe study the following performance metrics for

    DSR and FORP:

    (i) Percentage of successful route discovery ratio

    (expressed as percentage) of the number ofsuccessful route discovery attempts to the total

    number of route discovery attempts.

    (ii) Number of retransmitted messages the numberof messages received at all the nodes in the network

    per successful route discovery attempt, averaged

    over alls-dsessions for the entire simulation time.

    (iii)Number of retransmitting nodes the number of

    nodes retransmitting the RREQ packet in thenetwork per successful route discovery attempt,

    averaged over all s-d sessions for the entire

    simulation time.(iv) Number of route transitions average of the

    number of route discoveries required for all s-d

    sessions.

    (v)Hop count per route average of the number ofhops in routes, time-averaged over alls-dsessions.

    4.5 Percentage of Successful Route DiscoveryWe refer to a successful route discovery as the

    scenario when at least one RREQ packet broadcast

    by the source reaches the destination. The flooding,MPR and MCDS approaches guarantee successful

    route discovery if the underlying network is

    connected. With the probability and distance-based broadcasting techniques, there is always a chance

    that the RREQ packet does not reach its intended

    destination, even though the underlying network may

    be connected. The network density plays a huge rolein determining the minimum value for the probability

    of retransmission and the maximum threshold

    distance value for retransmission that wouldmaximize the number of successful route discoveries.

    Larger the network density, the lower the minimum

    probability of retransmission and larger the

    maximum threshold distance for retransmission thatwould maximize the chances of a successful routediscovery. In this paper, we set ourselves a target of

    92-95% successful route discoveries for each s-d

    session.For a given probability of retransmission, the

    percentage of successful route discoveries increases

    as the network density increases. For a givennetwork density, the percentage of success in route

    discoveries increases as the probability of aretransmission increases. At high network density,

    one can obtain 100% success in route discoveries

    when the probability of a retransmission is beyond0.7. With 25 nodes in the network, the maximum

    achievable percentage of successful route discovery

    is only 95%. Such a limitation arises due to the poor

    connectivity of the network at low density. For eachnetwork density, we define a Threshold Probability,

    ProbThresh, as the probability of retransmission that

    results in 92-95% success in route discoveries and atthe same time the number of retransmitted messages

    and the number of retransmitting nodes is the

    minimum. For fixed probability of retransmission

    values belowProbThresh, the percentage of success inroute discoveries decreases with increase in node

    mobility. This is due to the increase in the number of

    route discovery attempts as the node mobilityincreases. The value ofProbThresh was observed to be

    0.7 with a network of 25 nodes and 0.4 with a

    network of 50 nodes. The percentage of successful

    route discoveries for DSR under the probabilistic

    schemes is shown in Figures 1.1 and 1.2. Similarresults are obtained for FORP too.

    Figure 1.1: Network of 25 Nodes

    Figure 1.2: Network of 50 Nodes

    Figure 1: Percentage of Successful RouteDiscoveries with Probabilistic Scheme

    Similarly, to obtain 92-95% success in route

    discovery attempts, we choose DistTresh = 100m asthe maximum threshold distance value for

    retransmission in a network of 25 nodes and DistTresh

    = 180m as the maximum threshold value in anetwork of 50 nodes. The percentage of successful

  • 8/7/2019 Broadcast Strategies Meghanathan 76

    6/12

    route discoveries for DSR under the distance-basedscheme is shown is shown in Figures 2.1 and 2.2.

    Figure 2.1: Network of 25 Nodes

    Figure 2.2: Network of 50 Nodes

    Figure 2: Percentage of Successful Route

    Discoveries with Distance-based Scheme

    Figure 3 shows the percentage of successful route

    discoveries using the selected threshold values forthe probability and distance-based schemes and the

    other broadcasting techniques, including flooding.

    Figure 3: Percentage of Successful Route

    Discoveries with Different Broadcasting Techniques

    4.6 Reduction in Retransmission OverheadSince we did not let any intermediate node to

    reply for RREQ messages, the number of

    retransmitting nodes (Figure 4) and the number of

    retransmitted messages (Figure 5), depend only on

    the network density, node mobility and the

    broadcasting strategy used. With simple flooding

    (Figure 4), each node retransmits the RREQ messageexactly once. Hence, as the network density

    increases, the number of retransmitting nodesincreases. The destination node gets the RREQ

    message through several paths and thus can choose

    the best path depending on the route selection principles of the protocol employed. The route is

    learnt with the least possible route-acquisition delay,

    but with the maximum message retransmission

    overhead.On the other hand, the number of retransmitting

    nodes in the case of MCDS based route discovery is

    the minimum since the RREQ message isretransmitted only by nodes that are part of the

    approximate MCDS. But, the MCDS approach tends

    to increase the route-acquisition delay, as prior to

    route discovery, the MCDS itself needs to bedetermined. We run a distributed version of the

    Kous heuristic [13] in the network to approximate

    the MCDS. Each node then learns the set of itsMCDS neighbors and the presence/absence of the

    node in the MCDS.

    Figure 4: Average Number of Retransmitting Nodes

    per Route Discovery

    Figure 5: Average Number of Retransmitted

    Messages per Route Discovery

    For the MPR, the probabilistic and distance-based

    schemes, the number of retransmitting nodes and thenumber of retransmitted messages is in between the

    two extremes set by simple flooding and MCDS. TheMPR approach is not scalable as it does not take into

    account the path taken by the RREQ message. The

    set of MPR nodes is selected statically using agreedy approach of choosing neighbor nodes that

    covered the maximum number of 2-hop neighbors.

    When a node receives a RREQ message, the node

    does not remove from its MPR set the neighbor

  • 8/7/2019 Broadcast Strategies Meghanathan 76

    7/12

    nodes that might also have received the RREQmessage. The number of 1-hop and 2-hop neighbors

    of a node is doubled as the network density is

    doubled. As a result, the number of nodesconstituting the MPR set (the number of

    retransmitting nodes) also doubles, when the number

    of nodes in the network is increased from 25 to 50

    nodes.When simulated under the probabilistic and

    distance-based schemes using the threshold values

    mentioned in Section 4.5, we observe that thenumber of retransmitting nodes (Figure 4) required

    in a network of 50 nodes is only 20% more than the

    number of retransmitting nodes required in a network

    of 25 nodes. Similarly, we observe that in a low-density 25-node network operated atDistTresh = 100m,

    the number of retransmitting nodes required to

    guarantee a 92-95% success in route discovery isonly 40% more (and does not double) than that

    required in a high-density 50-node network operated

    atDistTresh = 180m.

    From Figure 5, we also observe that the number

    of retransmitted messages with flooding and MPRquadruples as we double the network density. This

    illustrates that flooding and MPR are not scalable

    broadcasting techniques. For the MCDS scheme, thenumber of retransmitted messages just doubles as the

    network density doubles. With the probabilistic

    scheme operated under the appropriate threshold

    values, the number of retransmitted messages in a50-node network is 2.4 times to that incurred in a 25-

    node network. With the distance-based schemeoperated under the appropriate threshold values, the

    number of retransmitted messages in a 50-node

    network is only 1.3 times to that incurred in a 25-

    node network. These two observations illustrate that

    the probabilistic and distance-based schemes, whenoperated at the appropriate threshold values for

    retransmissions, are scalable. This is one of the

    significant contribution and finding of our research.

    4.7 Average Hop Count per PathFrom Figures 6 and 7, one can observe that the

    average hop count per path for both DSR and FORP

    is not very much influenced by node mobility and is

    only affected by the broadcasting strategy used.When simple flooding is used as the route discovery

    strategy, the destination node learns about several

    routes from the source of the RREQ message to itself.

    From this set, the destination node can then choosethe best route according to the route selection

    principles of the routing protocol. When we employ

    the different broadcasting strategies, we are reducing

    the number of retransmitting nodes as well as thenumber of retransmitted messages. Hence, the

    destination node learns only relatively fewer routes

    compared to the situation when flooding is used.

    Figure 6.1: Network of 25 Nodes

    Figure 6.2: Network of 50 Nodes

    Figure 6: Average Hop Count per Path for DSR

    In the case of DSR (Figures 6.1 and 6.2), the hop

    count of the routes chosen using MCDS and floodingis the minimum. Routing through the nodes that form

    the minimum connected dominating set results in themessage traversing the minimum number of

    intermediate hops from the source node to the

    destination node. Figures 6.1 and 6.2 illustrate thatflooding also discovers a similar route with the

    minimum number of hops from the source node to

    the destination node. With MPR, probability-based

    and distance-based schemes, the hop count of DSRroutes increases by at most 15% compared to that

    discovered using MCDS and flooding.The hop count of FORP routes (Figures 7.1 and

    7.2) is normally more than that of DSR routes.

    Among a set of routes learnt, FORP selects the routethat has the largest predicted route expiration time.

    For such routes, at the time of their selection, the

    average physical distance of the constituent nodes ofa hop is only 55-65% of the transmission range of

    the nodes. This results in the relatively larger hop

    count for FORP routes.

    Figure 7.1: Network of 25 Nodes

  • 8/7/2019 Broadcast Strategies Meghanathan 76

    8/12

    Figure 7.2: Network of 50 Nodes

    Figure 7: Average Hop Count per Path for FORP

    The protocols learn the maximum and minimumnumber of source-destination (s-d) routes using

    flooding and MCDS respectively. Thus, the average

    hop count of FORP routes is 10-15% and 2-3% morethan that of DSR routes when the routes are learnt

    respectively using flooding and MCDS.

    In a probability-based scheme (Figures 8.1through 8.4), the number of retransmitting nodes

    decreases as the probability for retransmission isreduced. At high network density, the dense

    coverage of nodes within a neighborhood offsets for

    the lower threshold probability of retransmission. Atlow network densities, one has to adopt reasonably

    high values for the threshold probability of

    retransmission in order to guarantee a highpercentage of success in route discoveries.

    At high node mobility, the hop count of the

    routes decreases drastically as the probability for

    retransmission falls below 0.4 (for low densitynetworks) and 0.2 (for high density networks). This

    could be attributed to the loss of connectivity between the source and the destination for low

    values of the probability of retransmission. Thenetwork is partitioned into two or more segments.

    There exists a path from the source to the destination

    only if the two end nodes of the path are within thesame segment, thus accounting for the reduction in

    the hop count when the network is partitioned. As

    MPR incurs more message retransmissions, if we cantolerate a 15% sub-optimality in the hop count, the

    distance-based or probabilistic schemes, at the

    appropriate threshold values, may be preferred as the

    route discovering strategies for DSR.

    Figure 8.1:vmax = 5m/s, 25 Nodes

    Figure 8.2:vmax = 5m/s, 50 Nodes

    Figure 8.3:vmax = 50m/s, 25 Nodes

    Figure 8.4:vmax = 50m/s, 50 Nodes

    Figure 8: Probability of Retransmission Vs AverageHop Count per Path

    Figure 9.1:vmax = 5m/s, 25 Nodes

    Figure 9.2:vmax = 5m/s, 50 Nodes

  • 8/7/2019 Broadcast Strategies Meghanathan 76

    9/12

    Figure 9.3:vmax = 50m/s, 25 Nodes

    Figure 9.4:vmax = 50m/s, 50 Nodes

    Figure 9: Threshold Distance of Retransmission VsAverage Hop Count per Path

    In a distance-based scheme (Figures 9.1 through9.4), a node rebroadcasts the RREQ message only if

    no neighbor node within the area covered by the

    threshold distance of retransmission has yet

    broadcasted the message. In the case of DSR, even

    though we may use several threshold distance values

    for retransmission, the protocol chooses only theroute that has the minimum hop count. Hence, the

    hop count of DSR routes is not much sensitivetowards the threshold distance for retransmission,

    except for high values of the distance. FORP isslightly more sensitive to the threshold distance of

    retransmission. Routes with physical hop distance55-65% of the transmission range are more likely to

    be found when the threshold distance for

    retransmission of the RREQ messages is also only55-65% of the transmission range of the nodes.

    4.8 Average Number of Route TransitionsIn the case of DSR (Figure 10), routes discovered

    through flooding and MCDS have the minimum

    number of hops. But such routes are very unstable as

    observed in Figures 10.1 and 10.2. At the time of

    route discovery, the average physical distancebetween the constituent nodes of a hop is close to 70-

    80% of the transmission range of the nodes. Suchhops are highly vulnerable to fail as the constituent

    nodes of the hop are more susceptible to move away

    quickly. The chance of link failure in the near futureincreases with increase in node mobility.

    Broadcasting strategies like MPR also do not offer

    any improvement in the stability of the routes chosen.

    The DSR protocol always targets for the minimum

    hop route among the set of routes discovered usingthese broadcasting strategies. At the threshold values

    for the probability of retransmission and the

    threshold distance for retransmission, as indicated inFigures 10.1 and 10.2, DSR incurs 20% less

    transitions compared to routes discovered using

    flooding.

    Figure 10.1: Network of 25 Nodes

    Figure 10.2: Network of 50 Nodes

    Figure 10: Stability of DSR Routes

    Figure 11.1: Network of 25 Nodes

    Figure 11.2: Network of 50 Nodes

    Figure 11: Stability of FORP Routes

    In the case of FORP (Figures 11.1 and 11.2), the

  • 8/7/2019 Broadcast Strategies Meghanathan 76

    10/12

    routes are most stable when discovered usingflooding. This is because the targeted destination

    node of the RREQ message receives the message

    across several routes and selects the route with thehighest predicted route expiration time. When routes

    are discovered using flooding, the number of route

    transitions incurred by DSR is 70% and 125% more

    than that incurred by FORP routes at low and highnetwork densities respectively.

    When route discovery is done using MCDS, the

    RREQ messages are propagated only by the nodes inthe MCDS and hence, the routes learnt are very

    likely to be of minimum hop paths. Such routes are

    least stable. When routes are discovered using

    MCDS, the number of route transitions incurred byDSR is only 3-7% more than that incurred by FORP.

    Thus, FORP routes selected using MCDS based

    scheme are the most unstable of routes selected usingthe broadcasting strategies.

    DSR routes are less stable in networks of high

    density compared to networks of low density. This is

    due to the edge effect problem [14]. In high

    density networks, the average physical distance of ahop in a minimum-hop path during its discovery is

    close to 80% of the transmission range of the node.

    While in low-density networks, the average physicaldistance of a hop is only 70% of the transmission

    range of the nodes. In high-density networks, when

    we aim for minimum-hop, we can select the farthest

    neighbor that is on the path towards the destination.But, this results in routes that are highly unstable.

    When operated at the threshold distance forretransmission as shown in Figures 11.1 and 11.2,

    the number of route transitions incurred for both the

    protocols when using threshold distance of 180m is

    at most 1.5 times to that incurred when using

    threshold distance of 100m. More detailed results areshown in Figures 12.1 to 12.4.

    Figure 12.1:vmax = 5m/s, 25 Nodes

    Figure 12.2:vmax = 5m/s, 50 Nodes

    Figure 12.3:vmax = 50m/s, 25 Nodes

    Figure 12.4:vmax = 50m/s, 50 Nodes

    Figure 12: Threshold Distance of Retransmission VsAverage Number of Route Transitions

    Compared to the distance-based scheme, FORP

    routes discovered using MPR and probability-based

    scheme are relatively more stable. The number oftransitions incurred by these routes is only 20-35%

    more than that incurred by routes discovered usingflooding. For low density networks and in networks

    with high node mobility, the network connectivity is

    very limited when operated with low values for the probability of retransmission (Figures 13.1 through

    13.4). Under such conditions, the number ofsuccessful route discoveries and the number of route

    transitions are low.

    Figure 13.1:vmax = 5m/s, 25 Nodes

    Figure 13.2:vmax = 5m/s, 50 Nodes

  • 8/7/2019 Broadcast Strategies Meghanathan 76

    11/12

    Figure 13.3:vmax = 50m/s, 25 Nodes

    Figure 13.4:vmax = 50m/s, 50 Nodes

    Figure 13: Probability of Retransmission VsAverage Number of Route Transitions

    5 CONCLUSIONSThe high-level contribution of this paper is a

    simulation-based analysis on the impact of the

    broadcast route discovery techniques on the stability

    and hop count of routes discovered for the minimum-

    hop based DSR and the stability-based FORPprotocols. We also showed that the probability-based

    and distance-based schemes, when operated at theappropriate threshold values for retransmission, are

    more scalable compared to the flooding and MPRschemes. Future work will also involve studying the

    impact of the broadcasting strategies on the linkefficiency and stability of trees and meshes

    determined for the multicast routing protocols.

    For networks of low density and high density, weidentify the threshold values for the probability of

    retransmission and the distance for retransmission,

    using which we can obtain 92-95% success in route

    discoveries with the minimum number of

    retransmissions, and below these threshold values,the percentage of success in route discoveries

    decreases with increase in node mobility. When

    operated at the threshold probability values forretransmission, the number of retransmitting nodes in

    a network of 50 nodes is only 20% more than the

    number of retransmitting nodes in a network of 25

    nodes. Also, when operated at the appropriatethreshold distances for retransmission, the number of

    retransmitting nodes decreases as the network

    density increases. The probabilistic and distance- based schemes require less overhead to implement

    compared to the MPR and MCDS based schemes.

    Determining the MCDS in a highly mobile network

    itself will be a significant overhead.When we employ the different broadcasting

    strategies, we are reducing the number of

    retransmitting nodes as well as the number ofretransmitted messages. The routing protocols learn

    only relatively fewer routes compared to the situation

    when flooding is used. With flooding, each node in

    the network retransmits the RREQ packet exactlyonce, thus resulting in the maximum number of

    retransmissions. Letting the RREQs propagate

    through the nodes that form the minimum connecteddominating set results in the packet traversing the

    minimum number of intermediate hops with

    minimum number of retransmissions from the source

    to the destination. So, we learn the maximum andminimum number of routes using flooding and

    MCDS respectively. The number of routes learnt

    using the other broadcasting strategies is in betweenthese two extremes.

    In the case of DSR, the hop count of routes

    chosen using the MCDS and flooding based route

    discovery approaches is the minimum. Nevertheless,

    since DSR opts always for the minimum hop routes,the hop count of DSR routes discovered using MPR,

    probability-based and distance-based schemes is at

    most 15% more than that discovered using floodingand MCDS. This illustrates that routes having

    minimum hop or close to being minimum hop are

    very much discovered using the different

    broadcasting strategies. If we can tolerate a 15% sub-optimality in the hop count, the distance-based or

    probabilistic scheme at the appropriate thresholdvalues (which yield the minimum number of

    retransmissions) may be preferred as the route

    discovery strategies for DSR.

    FORP targets stable routes and the hop count of

    such routes are usually more than that of minimumhop routes. At the time of route selection, the

    average physical distance of the constituent nodes of

    a hop in stable routes is only 55-65% of thetransmission range of the nodes. Thus, FORP is more

    sensitive to the different broadcasting strategies. The

    average hop count of FORP routes is 10-15% more

    than that of DSR routes when routes are learnt usingflooding, MPR, distance-based and probabilistic

    approaches. While using MCDS, the hop count of

    FORP routes is only 2-3% more than that of DSRroutes.

    The stability of DSR routes does not change much

    with the broadcasting strategy used. This is because

    the protocol always targets for minimum-hop routesand manages to discover routes with minimum hopcount or routes close to minimum hop count

    irrespective of the broadcasting strategy used. With

    respect to FORP, the most stable routes arediscovered using flooding. FORP routes discovered

    using the MCDS approach are the least stable as they

    are more or less similar to DSR routes. FORP routesdiscovered using MPR and probability-based

  • 8/7/2019 Broadcast Strategies Meghanathan 76

    12/12

    schemes (operated at the threshold probability forretransmission) incur only 20-35% more transitions

    compared to those routes discovered using flooding.

    With regards to distance-based schemes, FORProutes are more stable when discovered using

    moderate values for the distance of retransmission.

    This is because, at the time of route discovery itself

    the physical distance between the constituent nodesof a hop is at least the threshold distance of

    retransmission. In general, route discoveries with

    less retransmission overhead yield less stable routesand vice-versa. We thus observe a tradeoff between

    the number of retransmissions per stable route

    discovery and the number of stable route discoveries

    needed for a source-destination session.

    6 REFERENCES[1] S. Ni, Y. Tseng, Y. Chen and J. Sheu: The

    Broadcast Storm Problem in a Mobile Ad Hoc

    Network, Proceedings of the 5th ACM International

    Conference on Mobile Computing and Networking,pp.151-162 (1999).

    [2] B. Williams and T. Camp: Comparison ofBroadcasting Techniques for Mobile Ad Hoc

    Networks,Proceedings of the 3rd

    ACM International

    Symposium on Mobile Ad Hoc Networking andComputing, pp. 194 205 (2002).

    [3] D. B. Johnson, D. A. Maltz and J. Broch: DSR:The Dynamic Source Routing Protocol for Multi-hop

    Wireless Ad hoc Networks, Ad hoc Networking,

    edited by Charles E. Perkins, Chapter 5, Addison

    Wesley, pp. 139 172 (2001).

    [4] W. Su, S-J Lee and M. Gerla: Mobility Prediction

    and Routing in Ad Hoc Wireless Networks,

    International Journal of Network Management, Vol.11, No. 1, pp. 3-30 (2001).

    [5] J. Broch, D. A. Maltz, D. B. Johnson, Y. C. Hu

    and J. Jetcheva: A Performance Comparison ofMulti-hop Wireless Ad Hoc Network Routing

    Protocols, Proceedings of the 4th ACM Annual

    International Conference on Mobile Computing andNetworking,pp.85 97 (1998).

    [6] P. Johansson, T. Larson, N. Hedmanm B.

    Mielczarek and M. DegerMark: Scenario-basedPerformance Analysis of Routing Protocols forMobile Ad Hoc Networks, Proceedings of the 5th

    ACM International Conference on Mobile

    Computing and Networking,pp. 195 206 (1999).

    [7] N. Meghanathan and A. Farago: Survey and

    Taxonomy of 55 Unicast Routing Protocols forMobile Ad Hoc Networks, Technical Report UTDCS-

    40-04, University of Texas at Dallas (2004).

    [8] N. Meghanathan: A Simulation Study on the

    Stability-Oriented Routing Protocols for Mobile AdHoc Networks, Proceedings of 3rd IEEE

    International Conference on Wireless and Optical

    Communications Networks (2006).

    [9] Kaplan ED (ed.), Understanding the GPS:

    Principles and Applications, Artech House: Boston,

    MA (1996).

    [10] K. Fall and K. Varadhan: The ns Manual, The

    VINT Project, A Collaboration between researchers

    at UC Berkeley, LBL, USC/ISI and Xerox PARC.

    [11] IEEE Standards Department, Wireless LANMedium Access Control (MAC) and Physical Layer

    (PHY) Specifications, IEEE Standard 802.11-1997

    (1997).

    [12] C. Bettstetter, H. Hartenstein and X. Perez-

    Costa: Stochastic Properties of the Random-WayPoint Mobility Model, Wireless Networks, Vol. 10,

    No. 5, pp. 555 567 (2004).

    [13] L. Kou, G. Markowsky and L. Berman: A Fast

    Algorithm for Steiner Trees, Acta Informatica, Vol.

    15, pp. 141 145 (1981).

    [14] G. Lim, K. Shin, S. Lee, H. Yoon and J. S. Ma:

    Link Stability and Route Lifetime in Ad hocWireless Networks, Proceedings of International

    Conference on Parallel Processing Workshops, pp.

    116 123 (2002).