Neighborhood-based Route Discovery Protocols ... - Semantic Scholar

Neighborhood-based Route Discovery Protocols for Mobile Ad hoc Networks Sanaa A. Alwidian1*†, Ismail M. Ababneh2‡, and Muneer O. Bani Yassein3 1

ᶲ

Department of Computer Science and Applications The Hashemite University, Zarqa 13115, Jordan 2

Department of Computer Science Al al-Bayt University, Mafraq 25113, Jordan 3

Department of Computer Science Jordan University of Science and Technology, Irbid 21100, Jordan †

E-mail: [email protected] E-mail: [email protected] ᶲ E-mail: [email protected] ‡

* Corresponding author. Tel: +962 5 390 3333 ext. 4786; E-mail: [email protected] Abstract: Network–wide broadcasting is used extensively in mobile ad hoc networks for route discovery and for disseminating data throughout the network. Flooding is a common approach to performing network-wide broadcasting. Although it is a simple mechanism that can achieve high delivery ratio, flooding consumes much of the communication bandwidth and causes serious packet redundancy, contention and collision. In this paper, we propose new broadcast schemes that reduce the overhead associated with flooding. In these schemes, a node selects a subset of its neighbors for forwarding the packet being broadcast to additional nodes. The selection process has for goal reducing the number of neighbors and maximizing the number of nodes that they can reach (i.e., forward the packet to). By applying this novel neighborhoodbased broadcasting strategy, we have come up with routing protocols that have very low overhead. These protocols were implemented and simulated within the GloMoSim 2.03 network simulator. The simulation experiments show that our routing protocols can reduce the overhead for both low and high mobility substantially, as compared with the well-known and promising AODV routing protocol. In addition, they outperform AODV by increasing the delivery ratio and decreasing the end-to-end delays of data packets. Keywords: MANET, AODV, Broadcasting, Flooding, Route Discovery

1. Introduction A Mobile Ad hoc Network (MANET) is an autonomous ad hoc network consisting of a collection of mobile nodes that utilize wireless transmission for communication and cooperation. MANETs are self-configured, self-organized and self-controlled, without reliance on any preexisting infrastructure or centralized access points. Therefore, they can be deployed anytime and anywhere. The numerous applications of MANETs include search and rescue operations, academic and industrial applications, and Personal Area Networks (PANs).

A node in a MANET is required to operate as a host as well as a router that can forward packets so that they can reach nodes that do not reside within the transmission range of the source node. The topology of MANETs is dynamic. Nodes are free to change their physical location by moving freely in all directions (Yao Yu et al., 2009 ). A Network Wide Broadcast (NWB) is a common operation that is used extensively in MANETs to discover routes and to disseminate data throughout the network. Flooding is a common operation that is used to perform NWB. Flooding refers to the process whereby a node rebroadcasts a packet when it receives it for the first time (Rogers et al., 2005). Although flooding is simple and can achieve delivery to a large percentage of nodes in the network, it has been shown to be expensive and wasteful; it consumes much of the communication bandwidth, wastes network resources and causes serious redundancy, contention and collision, which are collectively referred to as the broadcast storm problem (Ni et al., 2002). Many researchers have identified the disadvantages of flooding, and they suggested various solutions in the literature (Ni et al., 2002; Tseng et al., 2003). Several of these solutions use a fixed threshold value (Bani Yassein et al., 2009; Sasson et al., 2003). A node that receives a broadcast packet participates in the NWB only if some local measure meets the threshold value. Some other schemes build a virtual backbone whose task is to disseminate the broadcast packet throughout the network (Alzoubi et al., 2002; Clausen & Jacquet, 2003) . Only backbone members are responsible for broadcasting packets. This approach is vulnerable to transmission losses and poor robustness, measured in terms of achieved coverage in the presence of losses. The virtual backbone becomes disconnected when a node moves away from its neighbor or neighbors. Also, location-based schemes were proposed. An issue with these schemes is that they depend on node location information that is typically provided by additional equipment, such as GPS devices (Williams & Camp, 2002). The main goal of the protocols proposed in this paper is also to reduce the overhead resulting from flooding. However, the strategy we propose is based on selecting a subset of neighbors that can forward a broadcast packet to a large number of nodes. Our protocols do not require distance measurement or exact location determination devices. A forwarding node that receives the broadcast packet selects a subset of its neighbors based on their ability to reach additional nodes, and only the selected neighbors will continue the broadcasting process. To begin with, the source node selects its forwarding neighbors that will participate in the process. By applying this strategy, we have, in particular, come up with route discovery protocols that have very low overhead. Yet, they are able to adapt quickly to changes in the network topology, providing also high packet delivery ratio and low end-to-end delay. The proposed protocols have been implemented and simulated using the GloMoSim 2.03 network simulator. The rest of this paper is organized as follows. Section 2 contains a review of previous research work related to network-wide broadcasting. In Section 3, we present the proposed neighborhood-based schemes. In Section 4, we discuss the simulation environment, the simulation parameters and the various performance metrics that are measured in the simulations. In addition, the simulation results are presented and analyzed. Simulation results for larger area are presented in Section 5. Finally, in Section 6, we conclude this paper and provide directions for future work.

2. Related Work In Mobile Ad hoc Networks, NWB is used extensively for many purposes, including route discovery, address resolution and carrying out other network layer tasks (Rogers & AbuGhazaleh, 2005). For instance, reactive routing protocols such as AODV (Perkin et al., 1999) and DSR (Johnson, 1994) benefit from the information gathered while broadcasting route request packets in maintaining a route table at every node. However, due to the dynamic nature of MANETs, routes break often and routing protocols are required to update the route tables frequently, causing a large number of broadcast messages to be disseminated across the network. In the literature, there are many schemes proposed for broadcasting in MANETs. They have been classified into the following five categories: simple, probabilistic, counter-based, area-based and neighbor-knowledge-based flooding (Ni et al., 2002; Williams et al, 2002). In simple flooding (Ni et al., 2002), a node broadcasts a received packet provided that it did not broadcast it before. Packets received previously are discarded. In this naïve flooding, a node rebroadcasts a packet at most once. Thus, the total number of rebroadcasts is in Θ(N), where N is the number of nodes in the MANET (Zhang & Agrawal, 2005). Although simple flooding is a straightforward approach that aims to reach every node in the network, it consumes much of the communication bandwidth, wastes network resources and causes serious redundancy, contention and collision, which are referred to as the broadcast storm problem (Ni et al., 2002). Probabilistic schemes (Haas et al., 2002) have been proposed for broadcasting/multicasting in wired and wireless networks. In such schemes, and upon receiving a new broadcast packet, a mobile host rebroadcasts that packet according to a specific probability, P (Sasson et al., 2003). It is obvious that when P=1, the probability-based approaches become similar to simple flooding. The proposed probability-based schemes are differentiated based on the method used for determining the value of P. Sasson et al. (2003) have proposed a probabilistic approach so as to reduce the redundant transmission of packets encountered in simple flooding and alleviate the broadcast storm problem. In this approach, a broadcast probability, P, is assigned in advance. Upon receiving a packet, a node rebroadcasts that packet according to the specified probability. It is demonstrated in the literature that the best value of P is around 0.7 (Tseng et al., 2003). Probabilistic flooding achieves good results compared with simple flooding. It reduces transmission redundancy, while being able to reach a large percentage of nodes. However, this approach uses the same probability without taking the density of the node's neighborhood area into consideration. For example, if the node is in a dense area (i.e., has many neighbors), the packet can reach the same set of nodes many times, resulting in broadcast redundancy. The broadcast probability should be set low in nodes located in dense areas. On the other hand, if the transmitting node has a small number of neighbors (i.e., it is located in a sparse area), it is less likely that the broadcast packet will reach the hosts in the transmission area of the node, thus the broadcast probability should be high (Bani Yassein et al., 2005). Zhang et al. (Zhang & Agrawal, 2005) proposed a dynamic probabilistic scheme that combines both probability-based and counter-based approaches. In this approach, a counter is

maintained at each node for counting the number of times a packet has been received. The packet counter is used as density estimator, although it does not necessarily correspond to the exact number of neighbors. Indeed, some neighbors may have suppressed their rebroadcasts according to their local rebroadcast probability (Zhang & Agrawal, 2005). The probability P is increased if the value of the packet counter is low (or equivalently if the current node is located in a sparse neighborhood), and it is decreased if the value of a packet counter is high. Compared with the probabilistic approaches where P is fixed, this dynamic approach has achieved higher throughput since the total number of rebroadcasts is reduced. However, the decision to rebroadcast or not is made after some delay. Bani-Yassin et al. (2006) have proposed a dynamic probabilistic approach, where the broadcast probability, P, is dynamically adjusted based on network density. To adjust the probability, short HELLO packets are used to count one-hop neighbors. If the number of neighbors is high, this indicates that the node is in a dense area. Thus, the chance of receiving numerous rebroadcasts of the same packet is high, and the probability P is set low to avoid redundancy. On the other hand, if the number of neighbors is small, P is set high to increase the chance of reaching the neighbors. In counter-based approaches, a specific threshold value is used, and the mobile host rebroadcasts the packet only if the number of copies received by that host is less than the threshold value (Sasson et al. 2003). The counter-based schemes control flooding by inhibiting the rebroadcast of a message if it has been received more than a fixed number of times. It is assumed that additional node coverage is not significant if the threshold value is exceeded. Counter-based schemes can achieve high delivery ratio and throughput, but they suffer from relatively long delays. Area-based approaches (Williams & Camp, 2002) make use of geographical information that is provided by GPS devices or physical layer support. Such additional information is exploited in making broadcasting decisions that control flooding and reduce redundant rebroadcasts. The reliance on GPS and other location devices is a disadvantage of the area-based approaches. Neighbor-knowledge-based schemes make rebroadcast decisions depending on information on neighboring nodes obtained by exchanging HELLO messages. One example of such approaches is flooding with self-pruning (Peng & Lu, 2000). In this scheme, a 1-hop neighbor list is maintained at each host, and this list is added to every broadcast packet. Upon delivering a packet to the neighbors of a node, each neighbor compares its list of neighbors with the list recorded in the packet. A packet is rebroadcast if some neighbors of the receiving node are not included in the list recorded in the packet. An issue with this scheme is that redundancy is not avoided. Two or more neighbors may have a common neighbor that is not listed in the packet, and these neighbors will broadcast the packet so as to reach this neighbor.

3. The Proposed Schemes Our proposed broadcast schemes use a novel neighborhood-based approach for dynamically selecting the group of nodes that forward the broadcast message. The source node selects a

subset of its 1-hop neighbors for forwarding the broadcast packet, includes their addresses in the packet header, and broadcasts the packet. A node that receives a broadcast packet is a forwarding node if its address is included in the packet header. Otherwise, it drops the packet. Forwarding nodes repeat the same process carried by the source. The two broadcast schemes we propose differ in the method used for selecting forwarding nodes. In the first method, a number of 1-hop neighbors that have the largest number of neighbors are selected as forwarding nodes. In the second method, a subset of 1-hop neighbors that can reach all 2-hop neighbors forms the forwarding group. The two schemes are respectively referred to as the Broadcast-based K-Neighbor Scheme, and the Broadcast-based Covering Neighbors Scheme. Below, they are described within the context of on-demand route discovery. In this case, the packet being broadcast is the Route REQuest (RREQ) packet, and the goal is finding a path to a destination node. When a RREQ packet reaches its destination node, the destination sends a reply to the source of the request, and it does not forward the packet. Information on neighbors that is used in the proposed schemes is obtained via HELLO messages that are exchanged periodically, as in AODV. 3.1. Broadcast-based K-Neighbor Scheme (BKNS) This scheme is an on-demand, broadcast-based ad hoc route discovery protocol that is designed for MANETs. The main goal of this scheme is to control the flooding process by reducing redundant broadcasts, which reduces the routing overhead. To facilitate the understanding of BKNS, we present a human activity called “cooperative search for a fugitive”, and adopt its logic in BKNS. 3.1.1 The Cooperative Search for a Fugitive We are in a police station, S, in a city, and we want to look for a fugitive hiding in a house, D. We assume that we do not know where the house is, and city dwellers are very cooperative. We can start the search process at the police station by searching in K (e.g., 3) neighboring houses that have the largest number of neighbors. From each of these houses, we are guided by their inhabitants to up to K neighbors that they know have the largest number of neighbors. A condition is that we never continue the search process from a house reached previously. Using this search method, we may be unable to find the house D, although we will likely reach most city houses. For example, D may be located in a sparse section of the city. Upon failure, we can make a thorough (naïve flooding) search starting from the police station. 3.1.2 Implementation of BKNS BKNS is implemented based on the cooperative search described above, where S is the source of the broadcast packet, D is its destination, and the houses and neighboring houses represent MANET nodes at relevant time instances. Figure 1 represents the topology of a MANET at some time instance. In BKNS, each node maintains a parameter called the degree of the node, where the degree of node X, degree(X), is the number of neighbors of this node. The degree of a node is equal to the size of that node’s neighbor table, nbrTable. This table contains an entry for each neighbor from which a HELLO

message was received within the previous time-period, called the hello-interval. In Figure 1, degree (S) =3, degree(C) = 12, degree (B) = 5 and degree (A) = 2.

C

S A

B

Figure 1: A MANET topology Every hello-interval, each node broadcasts a HELLO message containing its address and degree. Upon receiving the HELLO message, a node updates its routing and neighbor tables, such that an entry will be added in both tables for the node that sent the HELLO message, if it is not already in the table of neighbors. At any time, the neighbor table of node X, nbrTable(X), will contain the addresses of all X’s 1-hop neighbors and their degrees. The neighbor table entries are sorted in the decreasing order of the degree field. When a source node S wishes to communicate with a destination D, and there is no known route to this destination, it prepares a RREQ message and selects the first K one-hop neighbors that have the largest degrees as forwarding nodes. Through simulation experiments, we have tried K = 1, 2, .., 8, and have found that choosing K = 4 as the maximum number of forwarding nodes achieves good results for the simulation parameters considered. However, using K = 4 does not achieve good results in environments with low density. In this case, the performance of BKNS becomes almost exactly the same as the performance of AODV, since almost all nodes will participate in the route discovery process. That is, the value of K should depend on node density. We have experimented with K being a fraction of the number of neighboring nodes, N. 𝑵 Extensive simulation empirical evidence shows that K= 𝟑 performs very well for various densities. In what follows, we limit ourselves to this variant of BKNS. After determining its K candidate neighbors, the source node appends their addresses to the RREQ message. Upon receiving the RREQ message, only those nodes whose addresses are among the K-neighbors’ addresses will process the message and rebroadcast it further, as shown in Figure 2. The scheme BKNS is shown below in Figure 3.

Figure 2: BKNS broadcast example

1. Periodically, every HELLO_INTERVAL, broadcast a HELLO message containing own 2. 3. 4. 5. 6. 7. 8. 9.

address and degree. On receiving a HELLO message : update nbrTable(X), so that it will contain for neighbors. maintain the nbrTable(X) entries sorted in decreasing order according to the degree field. if X needs to communicate with a destination D, the following actions take place: if a route exists to the destination use it. else prepare a RREQ message, select the first K neighbors, and appends their addresses to the RREQ message to be sent, where K=

𝑵 𝟑

.

10. Upon receiving an RREQ message, the following actions take place: 11. if the recipient node is the destination, respond to the source. 12. else 13. only those intermediate nodes whose addresses are in the RREQ message will process the RREQ and rebroadcast it further. 14. If no response is received from D and the number of RREQ retries have been exhausted 15. source sends a flooding RREQ packet.

Figure 3: BKNS implementation algorithm

3.2. Broadcast-based Covering Neighbors Scheme (BCNS) This is the second ad hoc, on-demand, broadcast-based scheme that we propose for controlling flooding and reducing its overhead. In BCNS, the forwarding one-hop neighbors of a particular node are selected such that they cover all of that node’s two-hop neighbors. For a node X, we refer to the set of X's one-hop neighbors that cover all of its two-hop neighbors as CoveringSet(X), and the set of the two-hop neighbors as SuperSet(X). An important aspect of constructing CoveringSet is to keep this set as small as possible. This is because the smaller the set, the less the overhead. Unfortunately, the task of selecting the optimal covering set with minimum size is an NP-hard problem (Wikipedia, Accessed June, 2009). Therefore, in our BCNS scheme, we propose a greedy algorithm as a heuristic for constructing the CoveringSet, as illustrated in the next subsection. The idea of our BCNS scheme is clarified in Figure 4. For a node to calculate its CoveringSet, it requires the set of its 1-hop neighbors and 2-hop neighbors. To obtain the list of 1-hop neighbors, we depend on periodic HELLO messages that are sent periodically (every HELLO_INTERVAL) by each of the nodes. To obtain the list of 2hop neighbors, a node sends a list of its own neighbors with the HELLO message it transmits periodically. The proposed BCNS scheme has been implemented using the algorithm shown in Figure 5a.

Figure 4: BCNS scheme concept

1. Periodically, every HELLO_INTERVAL, broadcast a HELLO message containing own address, degree and list of addresses of 1-hop neighbors. 2. On receiving a HELLO message at a node X: 3. Update nbrTable(X), so that it will contain . 4. Sort the contents of nbrTable(X) in the descending order of the degree field. 5. If X needs to communicate with a destination D, the following actions take place: 6. If a route exists to the destination. 7. use it 8. Else 9. find a subset of 1-hop neighbors that cover all 2-hop neighbors by applying the CoveringSet(Y) heuristic shown in Figure 6. 10. Prepare a RREQ message, and include the addresses of the nodes in the CoveringSet(X) in the RREQ message. 11. Upon receiving an RREQ message, the following actions take place: 12. If the recipient node is the destination. 13. done. 14. Else 15. only those intermediate nodes whose addresses are in the RREQ message will process the RREQ and rebroadcast it further. 16. If the destination is not found and the RREQ_RETRIES timer expires 17. source sends a flooding RREQ packet.

Figure 5a: BCNS implementation algorithm

3.3. Heuristic for Calculating the CoveringSet Let us refer to the set of all 1-hop neighbors of node Y as 1-hop(Y), and the set of all neighbors of 1-hop neighbors of Y as 2-hop(Y). In addition, let SuperSet(Y) denote the set of unique 2-hop neighbors of Y. We have: SuperSet(Y) =∪∀ 𝑖 ϵ 2−ℎ𝑜𝑝 𝑌 . The subset of 1-hop(Y) that covers 2-hop(Y) (i.e., CoveringSet(Y)) is computed by the algorithm shown in Figure 5b.

Input: nbrTable(Y) - neighbor table for node Y. Output: CoveringSet(Y). 1. If 2-hop(Y)==NULL 2. return (0) 3. Else 4. SuperSet(Y) =∪∀ 𝑖 ϵ 2−ℎ𝑜𝑝 𝑌 . 5. Initialize CoveringSet(Y) to . 6. For nodes in the sorted 1-hop(Y) list do: 7. check if the current node has a path to some nodes in SuperSet(Y) and add it to CoveringSet(Y). repeat until all nodes in the original SuperSet(Y) computed in 4 are covered by CoveringSet(Y). return (CoveringSet(Y)). Figure 5b: Covering set construction heuristic

4. Performance Analysis In this research, we evaluate the performance of two routing protocols that are based on the proposed route discovery schemes and compare them with AODV, where route discovery is based on flooding. We have implemented the proposed BKNS and BCNS route discovery schemes within the GloMoSim network simulator version 2.03 (Zeng et al., 1998). This simulator already contains an implementation of AODV. In our simulation experiments, we model a network of 10, 20, and 50 nodes. The nodes are placed randomly in a rectangular flat area. Two different areas of dimensions equal to 600 m × 600 m and 1000 m × 1500 m were considered. The goal of using the larger area is to experiment with longer paths (i.e., paths with more hops) and lower node densities. The network bandwidth is 2 Mbps and the medium access control (MAC) layer protocol is IEEE 802.11. For experiments that investigate the effect of speed, the maximum node velocities (MaxSpeed) considered are 1, 5, 10, 20 and 50 m/s. Node velocities are distributed uniformly over the interval [0, MaxSpeed]. Additional simulation parameters are shown in Table 1. Parameter values adopted in this work have been used in the literature (e.g., (Trung et al., 2007)). Each simulation run lasts for 300 seconds, and runs are repeated ten times with different random seeds.

Table 1: General simulation parameters Parameter Simulator Routing protocols evaluated Simulation time Number of nodes Simulation area Transmission range Movement model Traffic type Data payload Packet rate Link bandwidth

Value GloMoSim 2.03 AODV, BKNS, BCNS 300 s 10, 20, and 50 nodes 600 m  600 m or 1000 m  1500 m 250 m Random-waypoint Constant Bit Rate (CBR) 512 bytes/packet 1, 2, 4, 6, and 8 packets/s 2 Mbps

4.1. Performance Metrics In comparing the performance of the protocols considered in this paper, we have used several common performance metrics. These are the control overhead, packet delivery ratio, end-to-end delay, and saved rebroadcasts (Sun et al., 2008). Control overhead The control overhead (overhead, for short) represents the ratio of the number of control packets generated by the protocol to the number of data packets received by the destinations. It is computed as follows: 𝑂𝑣𝑒𝑟ℎ𝑒𝑎𝑑 =

𝑁𝑜. 𝑜𝑓 𝑐𝑜𝑛𝑡𝑟𝑜𝑙 𝑝𝑎𝑐𝑘𝑒𝑡𝑠 𝑠𝑒𝑛𝑡 𝑁𝑜. 𝑜𝑓 𝑑𝑎𝑡𝑎 𝑝𝑎𝑐𝑘𝑒𝑡𝑠 𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑑

Packet Delivery Ratio The Packet Delivery Ratio (PDR) is the ratio of the number of data packets received by destination nodes to those sent by the source nodes. The PDR is computed as follows: 𝑃𝑎𝑐𝑘𝑒𝑡 𝐷𝑒𝑙𝑖𝑣𝑒𝑟𝑦 𝑅𝑎𝑡𝑖𝑜 =

𝑁𝑜. 𝑜𝑓 𝑝𝑎𝑐𝑘𝑒𝑡𝑠 𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑑 𝑁𝑜. 𝑜𝑓 𝑝𝑎𝑐𝑘𝑒𝑡𝑠 𝑠𝑒𝑛𝑡

Average end-to-end delay This performance parameter represents the average delay between the time when the data packet originates at the source node and the time it reaches the destination node. Saved Rebroadcasts The saved rebroadcasts performance parameter represents the ratio of the number of route request (RREQ) packets retransmitted to the total number of route request (RREQ) packets received by the nodes (Hanashi et al., 2008). Let r be the number of RREQ packets that are

received by the nodes, and let t be the number of RREQ packets that they retransmit, the percentage of saved rebroadcasts is computed as follows: 𝑟−𝑡 𝑆𝑎𝑣𝑒𝑑 𝑅𝑒𝑏𝑟𝑜𝑎𝑑𝑐𝑎𝑠𝑡𝑠 = ∗ 100% 𝑟 4.2. Simulation Results and Analysis In the following subsections, we present and analyze the simulation results obtained for various input parameters. Unless it is specified otherwise, the results are shown for the 600 m × 600 m area. 4.2.1. Effects of Speed and Number of Traffic Generators: The purpose of the simulation experiments summarized in this subsection is to study the effect of the speed of nodes on the performance of the protocols using different numbers of constant bit rate (CBR) traffic generators. In these experiments, MaxSpeed is varied from 1m/s to 50 m/s. We have conducted many experiments for 10, 15 and 20 CBR traffic sources, where each source generates a traffic load of 4 packets/s. However, Due to space limitations, we will present the simulation results for 20 sources. In Figure 6, BKNS and BCNS generate substantially less control overhead than AODV. Also, the overhead of AODV increases substantially with the speed of nodes. However, the overhead of BKNS and BCNS is relatively stable. For the maximum speed of 1 m/s, BKNS and BCNS outperform AODV by 29% and 46%, respectively. For the maximum speed of 50 m/s, BKNS and BCNS outperform AODV by 70% and 72%, respectively. Overall, the simulation results show that as the number of sources increases from 10 to 15 to 20, the control overhead increases for all speed values, and both BKNS and BCNS outperform AODV significantly for all source numbers considered. When the number of sources increases, the probability that packets collide becomes larger, leading to higher route discovery overhead.

Figure 6: Overhead for 20 sources and a CBR of 4 packets/s

It can be seen in Figure 7 that BKNS and BCNS have slightly higher delivery ratios than AODV for all maximum speed values. When the maximum node speed is low (1 m/s), BKNS and BCNS outperform AODV by 1.3% and 2%, respectively. For the high maximum speed value of 50 m/s, the performance improvements are 1.7% and 2.4%, respectively. The results of the simulation experiments show that increasing the number of sources reduces the delivery ratio for all protocols. Reasons for this reduction are packet collisions and dropped packets. Overall, BCNS and BKNS slightly outperform AODV in terms of packet delivery ration for all numbers of sources considered (10, 20 and 50 sources).

Figure 7: Packet delivery ratio for 20 sources and a CBR of 4 packets/s

In Figure 8, BKNS and BCNS reduce the end-to-end delay by over 20% as compared with AODV for all maximum speed values. Furthermore, as the number of traffic sources increases from 10 to 15 to 20, the end-to-end delays increase for all protocols. As more data packets are generated per time unit when the number of sources increases, higher queuing delays can be expected. However, BKNS and BCNS still outperform AODV in terms of average end-to-end delay for all source numbers considered. Figure 9 depicts the Saved Rebroadcasts (SRB) achieved by our protocols in comparison with AODV when the number of sources again equals 20. Similar trends were obtained for 10 and 15 sources. As the figure shows clearly, our schemes outperform AODV in terms of avoiding redundant retransmissions of received packets. In addition, both BKNS and BCNS prove their stability and ability to save rebroadcasts even for high speed values, whereas the performance of AODV degrades significantly as the maximum node speed increases. On average, BKNS and BCNS save 90% and 97% of the rebroadcasts for all maximum speed values. However, AODV saves 60% of the rebroadcasts for the maximum speed of 1m/s, and saves only 33% for the maximum speed of 50 m/s.

Figure 8: End-to-end delay for 20 sources and a CBR of 4 packets/s

Figure 9: SRB for 20 sources and a CBR of 4 packets/s 4.2.2. Effects of Traffic Load and Number of Nodes: The purpose of the simulations presented in this subsection is to investigate the influence of varying the traffic load of the sources. For this purpose, we consider the source packet rates of 1, 2, 4, 6, and 8 packets/s, where the number of CBR generators is 10 and node speeds are uniformly distributed over the interval [0, 50 m/s]. When the total number of nodes is 10, we say that the network is sparse; in contrast, when this number is 50, we say that the network is dense. We have conducted many experiments for 10, 20 and 50 nodes. However, due to space limitations, we will not show the simulation results when the number of nodes is 20. Rather, we show simulation results for 10 (sparse network) and 50 (dense network) nodes so as to show the impact that the density of nodes has on overall performance.

4.2.2.1. Sparse Network Results Here, the number of nodes in the network is 10, and each node generates a traffic load that varies from 1 packet/s to 8 packets/s. In Figure 10, it is clear that our protocols outperform AODV in terms of routing overhead for all traffic load values. This is because they control flooding by selecting only a subset of nodes for forwarding packets. Figure 10 results show that for low traffic load, BKNS and BCNS outperform AODV by about 40% and 56%, respectively. When the traffic load is high, the improvement reaches 47% and 52% for BKNS and BCNS, respectively.

Figure 10: Overhead for 10 nodes and MaxSpeed = 50 m/s Figure 11 shows that BKNS and BCNS also outperform AODV in terms of PDR for all traffic load values. When the traffic load is low, both BKNS and BCNS outperform AODV by 7 percent. For the highest traffic load value, BKNS and BCNS outperform AODV by 12 and 14 percents, respectively.

Figure 11: Packet delivery ratio for 10 nodes and MaxSpeed = 50 m/s

The large overhead of AODV as compared with BKNS and BCNS increases packet end-toend delays. The average end-to-end delay of AODV is higher than that of the proposed protocols, as shown in Figure 12. The average end-to-end delay of AODV is larger than that of BKNS and BCNS by about 50 to 90 percent.

Figure 12: End-to-end delay for 10 nodes and MaxSpeed = 50 m/s Figure 13 shows that BKNS and BCNS outperform AODV in terms of SRB for all traffic loads considered. When the traffic load is 1 packet/s, BKNS and BCNS outperform AODV by 14 and 24 percents, respectively. For 8 packets/s, BKNS and BCNS outperform AODV by 15 and 26 percents, respectively.

Figure 13: SRB for 10 nodes and MaxSpeed = 50 m/s

4.2.2.2. Dense network Results The purpose of the simulation results presented in this subsection is to illustrate the impact of increasing the number of nodes on the performance of the protocols under study. It can be seen in Figure 14 that BKNS and BCNS outperform AODV in terms of control overhead for all source traffic load values considered. For the low source traffic load of 1 packet/s, BKNS and BCNS outperform AODV by 73% and 80%, respectively. For the source traffic load of 8 packets/s, the performance advantages of BKNS and BCNS are 62% and 84%, respectively. Comparing Figure 14 with Figure 10, it can be seen that the control overhead increases when the number of nodes is increased from 10 to 50. The reason is that more control packets are expected to be sent when there are more nodes in the network.

Figure 14: Overhead for 50 nodes and MaxSpeed = 50 m/s Figure 15 displays the packet delivery ratio obtained for the three protocols. It can be noticed in the figure that for low traffic loads (1 and 2 packets/s), our protocols and AODV have almost similar delivery ratio values. However, for high source traffic loads (6 and 8 packets/s), our protocols become substantially superior. This says that BKNS and BCNS are as effective as AODV in delivering packets to destinations for low traffic load values; however, they are substantially more effective for high traffic load values. Figure 16 displays the average end-to-end delays for the various source CBR traffic rates considered. The results show that as the traffic load increases, the average end-to-end delay for all protocols increases as well. Nevertheless, our protocols are still superior to AODV, especially for high traffic loads.

Figure 15: Packet delivery ratio for 50 nodes and MaxSpeed = 50 m/s

Figure 16: End-to-end delay for 50 nodes and MaxSpeed = 50 m/s

In Figure 17, we plot SRB against the source traffic load. The figure shows that BKNS and BCNS outperform AODV for all traffic loads. When the traffic load is low (1 packet/s), BKNS and BCNS outperform AODV by 61% and 64%, respectively. For the highest traffic load value considered (8 packets/s), BKNS and BCNS outperform AODV by 67% and 69%, respectively.

Figure 17: SRB for 50 nodes and MaxSpeed = 50 m/s

5. Simulation Results for Larger Area In this section, the main goal of the simulation experiments is to show the behavior of our protocols and the behavior of AODV when the simulation area becomes larger (i.e. 1000 m  1500 m) and MaxSpeed varies from the low speed of 1 m/s to the high speed of 50 m/s, where node speeds are again uniform over the interval [0, MaxSpeed]. The Figures 18-21 show consistency between the results of these experiments and those of the previous experiments. The simulation parameters are set as follows:     

Number of nodes: 50 nodes. Maximum speed: 1, 5, 10, 20, 50 m/s. Packet rate: 4 packets/s. Number of sources = 20 CBR generators. The other simulation parameters are set as in Table 1.

In Figure 18, it is clear that our protocols outperform AODV in terms of reducing the routing overhead for all speed values. This is because they control flooding by selecting only a subset of nodes for retransmitting packets. This reduces the number of control packets, which means a reduction in the overall routing overhead. Figure 18 shows that for the lowest speed value (MaxSpeed = 1 m/s), both BKNS and BCNS outperform AODV by 55%. When the maximum speed of nodes is high (50 m/s), the enhancement reaches 66% and 76% for BKNS and BCNS, respectively. Figure 19 results show that BKNS and BCNS outperform AODV for all speed values by a small percentage. When the maximum speed is lowest, BKNS and BCNS outperform AODV by 6 and 7.5 percents, respectively. For the highest maximum speed considered, BKNS and BCNS outperform AODV by 4 and 8 percents, respectively.

Figure 18: Overhead for 50 nodes, 20 sources, and CBR = 4 packets/s

Figure 19: Packet delivery ratio for 50 nodes, 20 sources, and CBR = 4 packets/s In Figure 20, the average end-to-end delay for all packets in AODV is higher than in our protocols. When the maximum speed of nodes is 1 m/s, AODV’s average end-to-end delay is more than twice that of BKNS and BCNS. Whereas, when the maximum speed of nodes is 50 m/sec, the enhancement of both BKNS and BCNS over AODV is about 33%. Figure 21 displays the SRB of the protocols and shows that BKNS and BCNS substantially outperform AODV for all speed values. When the speed is low (MaxSpeed = 1 m/sec), BKNS and BCNS outperform AODV by 58 and 59 percents, respectively. For the high maximum speed of 50 m/sec, BKNS and BCNS outperform AODV by 80 and 89 percents, respectively.

Figure 20: Average end-to-end delay for 50 nodes, 20 sources, and CBR = 4 packets/ second

Figure 21: SRB for 50 nodes, 20 sources, and CBR = 4 packets /second

6. Conclusions In this paper, we have proposed two neighborhood-based route discovery schemes for mobile ad hoc networks. The primary aim of these schemes is reducing the overhead associated with the route discovery process. The source node selects a subset of its one-hop neighbors for forwarding the route request packet further, and it includes their addresses in the request packet that it broadcasts. This process is repeated by every selected forwarding node, except the destination node. Non-forwarding nodes drop received route request packets. Thus, the forwarding nodes are selected dynamically in an expanding ring fashion starting with the source. At each step, the selection process has for goal reducing the number of upcoming forwarding nodes. Using extensive simulations, we have evaluated the proposed schemes and found that

they have very low overhead, yet they can achieve substantially higher delivery ratios than AODV when the traffic load is heavy. Even under moderate loads, they can achieve slightly higher delivery ratios than AODV, which is a successful and well-known routing scheme for ad hoc networks.

References Abdalla M. Hanashi, Aamir Siddique, Irfan Awan & Mike Woodward (2008), Dynamic Probabilistic Flooding Performance Evaluation of On-Demand Routing Protocols in MANETs. Proc. of the 2008 International Conference on Complex, Intelligent and Software Intensive Systems, Vol. 0. pp. 200-204. Baolin Sun, Chao Gui, Qifei Zhang, Bing Yan, and Wei Liu (2008), A Multipath On-Demand Routing with Path Selection Entropy for Ad Hoc Networks.Proc. of the 9th International Conference for Young Computer Scientists, Nov. 2008, pp. 558-563. Brad Williams &Tracy Camp (2002), Comparison of broadcasting techniques for mobile ad hoc networks. Proc. of the 3rd ACM International Symposium on Mobile Ad Hoc Networking and Computing (MOBIHOC 2002), ACM New York, NY, USA, pp. 194-205. Charles E. Perkin & Elizabeth M. Royer (1999), Ad-hoc on-demand distance vector routing.Proc. of the Second IEEE Workshop on Mobile Computing Systems and Applications (WMCSA '99), 25-26 Feb. 1999, pp. 90-100. David B. Johnson (1994), Routing in Ad Hoc Networks of Mobile Hosts. Proc. of the Workshop on Mobile Computing Systems and Applications, IEEE Computer Society, Santa Cruz, CA, Dec. 1994 , pp. 158-163. H. Trung, W. Benjapolakul& P. Duc(2007), Performance evaluation and comparison of different ad hoc routing protocols. Computer Communications, 30(11,12), 2478-2496. Khaled Alzoubi, Xiang-Yang Li, Yu Wang, Peng-Jun Wan & Ophir Frieder (2002) Messageoptimal connected dominating sets in mobile ad hoc networks. Proc. of the 3rd ACM international symposium on Mobile ad hoc networking and computing (MOBIHOC'02), 157– 164. M. Bani Yassein, A. Al-Dubai, M. Ould Khaoua & Omar M. Al-jarrah (2009), New Adaptive Counter Based Broadcast Using Neighborhood Information in MANETS. Proc. 2009 IEEE International Symposium on Parallel & Distributed Processing , May 2009, pp. 1-7. Muneer Masadeh Bani Yassein, Mohamed Ould-Khaoua & Stylianos Papanastasiou (2005), Performance Evaluation of Flooding in MANETs in the presence of Multi-Broadcast Traffic. Proc. of the 11th International Conference on Parallel and Distributed Systems (ICPADS05), vol. 2, July 2005 pages 505 – 509. Muneer Masadeh Bani Yassein, Mohamed Ould-Khaoua, Lewis M. Mackenzie, Stylianos Papanastasiou & A. Jamal (2006), Improving route discovery in on-demand routing protocols using local topology information in MANETs.Proc. of the ACM International Workshop on

Performance Monitoring, Measurement, and Evaluation of Heterogeneous Wireless and Wired Networks, ACM Press, Terromolinos, Spain, pp. 95–99. Muneer Masadeh Bani-Yassein, Mohamed Ould-Khaoua, Lewis M. Mackenzie & Stylianos Papanastasiou (2006), Performance analysis of adjusted probabilistic broadcasting in mobile ad hoc networks. International Journal of Wireless Information Networks, 13(2), 127–140. Paul Rogers, & Nael Abu-Ghazaleh (2005), Robustness of network-wide broadcasts in MANETs. Proc. of the 2nd IEEE International Conference on Mobile Ad Hoc and Sensor Systems (MASS 2005), Vol. 3, Nov. 2005, pp. 161-185. Qi Zhang & Dharma P. Agrawal (2005), Dynamic Probabilistic Broadcasting in MANETs. Journal of Parallel and Distributed Computing, vol. 65, pp. 220-233. Sze-Yao Ni, Yu-Chee Tseng, Yuh-Shyan Chen, & Jang-Ping Sheu (2002), The broadcast storm problem in a mobile ad hoc network. Wireless Networks, 8 (2), pp 153-167. Thomas Clausen &Philippe Jacquet (2003), Optimized link state routing protocol, Internet Draft, Internet Engineering Task Force, Oct. 2003, http://www.ietf.org/internet-drafts/draft-ietf-manetolsr-11.txt. Wei Peng, and Xi-Cheng Lu (2000), On the reduction of broadcast redundancy in mobile ad hoc networks. Proc. of the ACM International Symposium on Mobile Ad Hoc Networking & Computing (MOBIHOC), Boston, MA, 129-130. Wikipedia, the free encyclopedia, Online, Accessed June, 2009, "Set Cover Problem," Available from URL http://en.wikipedia.org/wiki/Set_cover_problem X. Zeng, R. Bagrodia & M. Gerla (1998), GloMoSim: a library for parallel simulation of largescale wireless networks.Proc. of the 1998 12th Workshop on Parallel and Distributed Simulations (PADS’98), Banff, Alb., Canada, May 26-29, 1998, pp. 154–161. Y.-C. Tseng, S.-Y. Ni &E.-Y. Shih (2003), Adaptive approaches to relieving broadcast storm in a wireless multihop mobile ad hoc network. IEEE Transactions on Computers, 52(5). Yao Yu, Yu Zhou, and Sidan Du (2009), Service Discovery in Mobile Ad Hoc Networks Using Mobility-Aware Attenuated Bloom Filters. Proc. 2009 IITA International Conference on Services Science, Management and Engineering, (July 2009), pp. 266-269. Yoav Sasson, David Cavin & André Schiper (2003), Probabilistic broadcast for flooding in wireless mobile ad hoc networks. Proc. of the IEEE Wireless Communications & Networking Conference (WCNC 2003), pp. 1124–1130. Yu-Chee Tseng, Sze-Yao Ni & En-Yu Shih (2003), Adaptive Approaches to Relieving Broadcast Storms in a Wireless Multihop Mobile Ad Hoc Network. IEEE Transactions on Computers, 52(5), pp. 545-557.

Zygmunt J. Haas, Joseph Y. Halpern, and Li Li (2002), Gossip-based ad hoc routing. Proc. of the IEEE INFOCOM, IEEE Computer Society, New York.