ABSTRACT

Mobile Ad Hoc Network (MANET) is a reliable system formed by co-operating and independent nodes that connect and communicate with each other wirelessly without pre-existing infrastructure. In such a network, the easiest way to broadcast the packets to all network nodes is flooding. However, flooding leads to serious drawbacks such as collision, contention and redundant retransmission. These drawbacks are known as the broadcast storm problem. This research proposes an adaptive algorithm in order to decrease the overall network overhead and minimize the problems associated to flooding. Moreover this research concentrates on studying the effect of using different mobility models that provide general idea about nodes movement, status and locations. In particular the performance of Random Waypoint and File mobility models in terms of their effect on network communication is evaluated. Moreover, this research provides performance evaluation of different traffic types such as Constant Bit Rate (CBR) and Traffic GEN. Results from Qualnet simulations have revealed that the new proposed algorithm outperforms flooding in terms of different metrics: reducing end-to-end delay in addition to minimizing the normalized routing load and maximizing the packets delivery ratio, in addition to concluding that Traffic GEN outperforms the CBR in terms of throughput. Moreover the CBR traffic has higher delay values than Traffic GEN.

1. Introduction

MANETs are infrastructure-less networks that can be deployed in short time. The easy deployment of nodes and self-organizing nature make MANETs suitable for a broad range of applications. Despite of the advantages of MANET, there are also multiple drawbacks in terms of power limitations, bandwidth limitations and dynamically changing topology [1] with high nodes mobility.

In such networks, mobile nodes function not only as a host but also as a router. The nodes movement occurs according to pre-determined parameters that define the nodes speed and nodes pause time. The packets transmission process across the network should be accomplished in such a way that consumes the minimum level of overall network resources and ensures the high level of packets delivery ratio. Flooding is the easiest transmission way that achieves high delivery ratio. In the other hand, flooding generates serious drawbacks by consuming network resources and bandwidth in addition to producing the well-known “broadcast storm problem”

which consists of redundant transmission for the packets, collision and contention.

This research concentrates on providing more effective transmission mechanism by using an adaptive algorithm that aims to alleviate the problems associated with flooding. This is achieved by using a counter-based broadcasting technique with a dynamic threshold value in order to obtain a high percentage of message reachability to network devices. The algorithm is proposed on the basis of initiating a counter c that is used to count the number of times; a node receives duplicate message and increment that counter by one value for each duplicate received packet. The adaptive counter based on scheme is used to minimize number of redundant broadcast messages, contention and collision by allowing only specific nodes to participate on broadcast propagation.

The arrangement of the nodes in such a network changes dynamically and in unexpected manner. Therefore, the mobility issue in MANET is a challenge and does not have a straight forward clarification. Due to that issue, various protocols had contributed to Ad Hoc networks. This paper also concentrates on evaluating the effect of using different mobility models.

The mobility models that are used widely in MANETs are mainly categorized into two types [2]: group-based mobility model, individual-based mobility model. There are different mobility models that can be easily applied to MANET with different applications, such as Random Waypoint, Random Walk, Group, and File mobility models.

Ad Hoc On-Demand Distance Vector Routing Protocol (AODV) [3] is well known as reactive protocol performing well in a wide variety of network topologies especially for those networks that are full of mobility such as MANETs. AODV aims to maintain a path towards the destination. Therefore, AODV reduces the overall network traffic by initiating on-demand routing requests. In such protocol there is no need to maintain full information about all nodes in advance, so the routes are discovered whenever needed through initiating route discovery process.

The main characteristics of AODV routing protocol are [3-5] reducing the overall processing overhead, providing effective maintenance capabilities for dynamic topologies, hop-by-hop routing, and loop-prevention capabilities. AODV performs well in order to find a path toward the destination by storing only the next neighbors’ information in the routing table. Therefore, AODV has evolved in order to control network performance and its scalability. There are different types of messages used by AODV in order to discover and maintain links. Whenever a source node decides to communicate with a specific destination, firstly the source node checks the routing table for the information related to the path towards that destination. The routing table either contains the desired path or not. In case the path is found, the source node transmits the data to the desired destination. Otherwise, the source node initiates the route discovery process. In order to perform that process, the source node immediately broadcasts a route discovery packet. Once the Route Discovery packet reaches the desired destination or reaches an intermediate node forward the packet to its destination, the route towards the destination is found. Therefore, the destination initiates a Route Replay packet towards the originator source either immediately or via intermediate node.

This research evaluated the performance of AODV protocol under two different mobility models [2,6-8]: Random Waypoint and File Mobility Model. We have analyzed the performance of each mobility model with respect to varying number of nodes. The movement of nodes is characterized according to its pause time, minimum and maximum speed. The nodes are uniformly distributed between the speed 0 and the maximum speed.

Random waypoint Model (RWP): In this model, the nodes are placed randomly within the simulation area. Each node in the simulation area starts moving from its original location toward a random location. Whenever the node reaches the target location, it waits for a duration known as “pause time” that is already determined in the mobility model parameters, and then the node moves to another target location. RWP model keeps on repeating the above procedure during the simulation time.

File Mobility Model: In this model, the nodes movement occurs according to a file specifying the next position for the nodes, in addition to specifying the time for nodes arriving to their locations. The movement of nodes occurs in a straight path with constant speed. However, in terms of mobile systems, this type of mobility models performs well in terms of time arrival expectation for the nodes, while all nodes movements are completely defined. In the other hand, it’s not widely applicable for highly randomized systems in which the destination positions are not highly known.

The rest of this paper is organized as follows: Section 2 introduces the related work studied in the literature. Section 3 describes simulation setup and performance metrics that we used to evaluate the performance of the proposed algorithm. Section 4 shows the result. Section 5 concludes the work.

2. Related Work

Zhang Qi et al. [9] improves the broadcasting mechanism by choosing P depends on the node density either its dense or sparse area. For example, if the node exists in dense area the broadcast probability should be relatively low in order to minimize the number of received duplicate packets. On other hand, if the node exists in a sparse area the broadcast probability should be relatively high in order to maximize packet delivery ratio to all nodes resides in the sparse area. In this approach the number of received packets used as an indicator of node density.

Muneer et al. [10] examined the performance of an adaptive counter-based scheme with adaptive threshold value in order decrease the drawbacks of simple flooding and increase the reachability percentage gained in the previous work. In this approach, the algorithm proposed to dynamically adjust an adaptive threshold value with three counters in order to gain more successful transmission with minimum rate of delay and overhead.

Harminder Bindra et al. [11] studied the performance of different mobility models over AODV and DSR routing protocols. This study revealed that, AODV outperforms DSR in most cases when using Constant bit rate (CBR) traffic. However, the DSR performs better when using TCP traffic in some situation (high mobility or high load).

Ajay Singh et al. [12] analyzed the performance of AODV routing protocol under two different mobility models: Random direction and Random walk. The authors studied the performance of the mentioned models in terms of different metrics: routing overhead, packet delivery ratio, average end-to-end delay, and normalized routing load. The simulation result revealed that, the Random Direction has a good result regarding packet delivery ratio and less routing load when the number of nodes increased.

Shailender Gupta et al. [13] evaluated the effect of using different mobility models such as Random waypoint, File, and Group mobility models on the performance of different routing protocols on MANET. This paper concluded that, every mobility model affects the overall performance of each protocol according to multiple parameters. Christian Bettstetteret et al. [14] analyze the performance of using Random waypoint mobility model on nodes distribution and status.

3. The Proposed Scheme

Now we will shed a light on the algorithm that proposed to deal with the broadcast storm problem. That’s where the proposed adaptive counter based scheme algorithm came up with an effective solution to network resources consumption and enhance the transmission process by allow subset of nodes participating in the packets forwarding process. Therefore, the remaining network nodes will stay keeping their resources such as batteries, energies and not spend much time to forward the already forwarded packets. Since the nodes that participate in the forwarding process will transmit the packet just in case the number of received duplicate packets is less than predefined threshold value. However, in order to perform that condition accurately, we have taking the network density into consideration either its sparse, dense or extra dense areas.

The network density effectively determined by exchanging the “Hello” control packets. The Adaptive Counter-based Broadcasting Scheme based on idea of initiating a counter c that keeps track of the received packets and therefore counts their number. Another counter threshold defined according to the node neighboring status. Since the nodes located in a sparse area has different threshold value than the nodes located in medium, dense or even extra dense area.

The threshold values c1, c2, c3 and c4 assigned to the sparse, medium, dense and extra dense area respectively. After determine those threshold values according to the neighboring information, the comparison process will take place between the counter c and one of those threshold values. Whenever the counter c is less than the threshold value, the broadcasting will continue. Otherwise the rebroadcast will inhibit.

The proposed algorithm performs as follows: whenever a node X heard a broadcast packet m, the node rebroadcast that packet if it’s received for the first time in addition to taking the node density into consideration as

the description bellow where n1, n2 and n3 is predefined values and c1 < c2 < c3 < c4:

• The node X rebroadcasts the packet according to c1 if the node located in sparse area, which means its neighbor numbers is less than or equal the average minimum number of neighbors n1.

• The node X rebroadcasts the packet according to c2 if the node located in medium area, which means its neighbor numbers is greater the minimum number of neighbors n1, and less or equal than the maximum numbers of neighbors n2.

• The node X rebroadcasts the packet according to c3 if the node located in dense area, which means its neighbor numbers is greater the maximum numbers of neighbors n2, and less than or equal the extra maximum numbers of neighbors n3.

• The node X rebroadcasts the packet according to c4 if the node located in extra dense area, which means its neighbor numbers is greater or equal the maximum numbers of neighbors n3.

• Main broadcast function is to deal with a specific packet and decide to rebroadcast it or not according to neighborhood information.

4. Simulation Setup

4.1. Simulation Tool Used

In order to evaluate the performance of the proposed algorithm and the two mobility models mentioned above, we compared them using Qualnet simulator. Qualnet simulator provides high scalability, accuracy, speed of simulation and result extracting.

4.2. Performance Metric Used

This work mainly focuses on a various metrics used as follows:

Average End-to-End Delay: measures the average time of delay taken by the successful received packets. This metric includes the overall delay during the route discovery, queuing, and retransmissions.

• Packet Delivery Ratio: measures the ratio of Total Packets Received by the destination to the Total number of Packets Sent by the source.

• Routing Overhead: Defined as the total number of Request Packets initiated and sent by the source during the simulation time.

• Throughput: Defined as the average rate of successful message delivered to the target location over a communication channel.

4.3. Simulation Parameter

Table 1 shows the various parameters used in the simulation process.