Performance of New Broadcast Forwarding Criteria in MANET

Performance of New Broadcast Forwarding Criteria in MANET Lijuan Zhu1 , Bu-Sung Lee1 , Boon-Chong Seet2 , Kai-Juan Wong3 , Genping Liu1 , Shell-Ying Huang1 , and Keok-Kee Lee1 1

Centre for Multimedia and Network Technology, School of Computer Engineering 2 Network Technology Research Centre Research TechnoPlaza, 4th Storey, Nanyang Technological University, Nanyang Avenue, Singapore 639798, 3

Institute for Computing Systems Architecture Informatics University of Edinburgh JCMB, Mayfield Road U.K, Edinburgh, EH9 3JZ

Abstract. In a mobile ad hoc network (MANET), packet broadcast is common and frequently used to disseminate information. Broadcast consume large amount of bandwidth resource, which is scarce in MANET environment. The problem is to find the ideal forward node set so as to minimize the bandwidth consumed, which is a NP-Complete problem. In this paper we will investigate three dominant node approximation algorithms: dominant pruning, total dominant pruning and partial dominant pruning algorithm. An extension to the original algorithms, modified termination criteria, is proposed. Simulation results show that the new termination criteria ensure the same coverage using a reduced number of forward nodes.

1

Introduction

A mobile ad hoc network (MANET) is a self-constructing network that consists of mobile hosts roaming around and communicating with each other freely. Due to the radio transmission range limit, packets transmitted from the source may need intermediate hosts to help relaying before they can reach the destination. Such network finds applicability in military environment, wherein a platoon of soldiers may establish an ad hoc network in the region of their deployment. It has been used as well in non-military environment, e.g. Inter-vehicular communication. There are several types of message communication services in ad hoc network. When one packet needs to be sent from one node to all other nodes in the network, a broadcast service is needed. The straightforward approach for broadcast is flooding, in which each node retransmits the received packet once. Many MANET routing protocols such as Ad Hoc On Demand Distance Vector (AODV) [1], Dynamic Source Routing (DSR) [2], Location Aided Routing (LAR) [3] and Zone Routing protocol (ZRP) [4] use flooding or its derivation to establish routes. This traditional flooding causes excessive redundant retransmissions, contention and congestion, which is referred as broadcast storm problem.

2

Some related works have been done to reduce redundant broadcast. Multipoint Relaying [6] restricts the number of neighbor nodes, which can reduce the retransmission by efficiently selecting a small set of neighbor nodes, which can cover the same region as the whole. This is a distributed mechanism; each node independently selects its own multipoint relaying set without the information of others’ selection. In [7] Jie Wu and Wei Lou used a greedy algorithm to select the forward node set at each node, and then broadcast in clustered networks based on forward node set. H. Lim and C. Kim proposed a dominant pruning algorithm [8], and later, in [9] Wei Lou and Jie Wu proposed other two algorithms: total dominant pruning algorithm and partial dominant pruning algorithm, which reduced some of the redundancy in the dominant pruning algorithm. They also discussed two termination criteria: marked and relayed. In this paper, we modify the termination criteria, and proposed a hybrid termination criterion that shows marked improvement in the reduction of number of forwarding nodes. The rest of this paper is organized as follows: section 2 introduces some graph definitions and forward node selection algorithms. Section 3 proposes the modification of the marked /relayed mechanism and the hybrid termination criteria, and examples are given at section 4. Section 5 shows the simulation results. Finally section 6 concludes the paper.

2

Problem Definition

An ad hoc network can be represented by a graph, G = (V, E), where V is the set of nodes in the network, and E is the set of edges between every two nodes. An edge exists between two nodes only when they are within the transmission range of each other, and then each one is called one-hop neighbor by the other node. For a node v, we use N (v) to represent its one-hop neighbor set (including v), and N (N (v)) to represent its two-hop neighbor set (the union of the N (v)’s one-hop neighbor). The hosts can get the one-hop or two-hop neighborhood information by periodically exchanging their “Hello” messages. The ideal way for broadcast is to select the minimum connected dominating set (MCDS) [5] nodes to do the rebroadcast. Finding MCDS is a NP-Complete problem, and extensive work has been done in the theoretical community on finding a good approximation of MCDS. The author in [5] proposed an approximation algorithm (AMCDS), which assumes that it has full network connectivity information. AMCDS Process: 1. Color all the nodes in V white. 2. Select the node with the maximum node degree, and put it into the set C. color this node black and all its neighbors white. 3. Select the grey node with the maximum white node degree, and put it into the set C. Color this selected grey node black and all its white neighbors grey. 4. If there are still some nodes white, go to 3; else go to 5.

3

5. The set C is an approximation for the MCDS. However, in the real MANET environment the nodes will only have limited information, eg. 1-hop or 2-hop neighbor information. Distributed algorithms [8][9] that make use of 1-hop and/or 2-hop information, has been proposed to find the forwarding node set in the MANET environment. These algorithms have two major tasks: selection of forwarding nodes and termination of broadcast. 2.1

Forward node selection algorithms

Three distributed algorithms for forward node selection were investigated. In the discussion that follows, the following parameters are defined: – N (v) is defined as the 1-hop neighbors of node v. – N (N (v)) is defined as the 2-hop neighbors of node v. – B(u, v) = N (v) − N (u). This is the set of nodes in the neighborhood of node v that are not the 1-hop neighbors of node u. – Ux (u, v): the set of forwarding nodes that would relay the packet based on the algorithm X. – F (u, v): the forwarding node set, initialized to zero members. – Z: initialized to zero members (empty set). The general algorithm for forward node selection when node v receives a broadcast packet from node u is as follows: 1. 2. 3. 4.

For every node wi ∈B(u, v), Si = N (wi )∩U (u, v) Find wj with the maximum size of its corresponding set Si . Add wj to F (u, v), Z = Z∪Sj , and for all the Si , Si = Si − Sj . If no new node is added to Z, exit; otherwise return to step 2.

Dominant Pruning (DP) Algorithm When node v receives a broadcast packet from node u, it will do the following steps to select the minimum forwarding nodes based on the process described above. For DP [7] the set of nodes that should be considered for forwarding the packet are as follows: UDP (u, v) = N (N (v)) − N (v) − N (u).

(1)

Thus, they are the nodes that are 2-hops away from node v, which are not member of the nodes that are 1-hop away from node v and node u. UDP (u, v) is shown as shaded areas in Figure 1. The two nodes u and v are marked as black dots. Total Dominant Pruning (TDP) Algorithm In TDP [8] the forwarding nodes set is given by: UT DP (u, v) = N (N (v)) − N (N (u)).

(2)

4

Fig. 1. Elimination of neighbor in DP

TDP makes full use of the 2-hop neighbor information in the selection of the forwarding nodes. Figure 2 shows the coverage of the neighbor nodes for TDP. The shaded region represents UT DP (u, v), the neighbor nodes area that would act as the broadcast forwarding nodes. The area is less than that of DP, thus less number of forwarding nodes. However, it suffers from the need for node u to piggyback its 2-hop neighbor set along with the broadcast packet.

Fig. 2. Elimination of neighbor in TDP

Partial Dominant Pruning (PDP) Algorithm In PDP [8], they remove the nodes that are 1-hop neighbors of node u and node v as well as the 2-hop neighbors of the nodes that are 1-hop neighbors of both node u and v. Let P = N (N (v)N (u)). (3) then UP DP (u, v) = N (N (v)) − N (v) − N (u) − P.

(4)

The shaded areas in Figure 3 represent the area where the forwarding nodes can reside as defined by equation 4, i.e. UT DP (u, v). The area covered is slightly larger than TDP. The advantage of PDP over TDP is that it does not need to

5

Fig. 3. Elimination of neighbor in PDP

piggyback the 2-hop neighbor set information of the sender with the broadcast packet. Thus, reducing the overhead. 2.2

Termination criteria

When a broadcast packet is sent, every intermediate node will use the DP/PDP/TDP algorithm to select the forward nodes, and every selected forward node will use termination criteria to determine whether to broadcast the packet. Wei Lou and Jie Wu [9] proposed two termination criteria: – Marked: When node v receives a broadcast packet it will not rebroadcast if all its one-hop neighbors’ status are marked. – Relayed : When node v receives a broadcast packet it will not rebroadcast if it has previously relayed the packet.

3 3.1

Modified Marked/Relayed Modified Marked

We modify the marked termination criterion described above, and propose a hybrid one based on the two termination criteria, which is as follows: When node v is performing its forward node selection process, it will drop its marked neighbors out of consideration. Node v will stop rebroadcast only if all its one-hop neighbors’ status are marked. Theorem 1. If node v is marked, then its one-hop neighbor N (v) has been marked before or will be marked later. Proof. Node v is marked, assume that it receives the packet from its neighbor u, there are then two conditions: (1) u will select v as the forward node, if v really broadcasts the packet using the termination criteria, then its one-hop neighbor N (v) will receive that packet and be marked later, otherwise its N (v) has been marked before. The theorem is thus correct; (2) u will not select v as the forward node, then u will select other forward nodes to mark N (N (u)), N (v)⊆N (N (u)), so N (v) will be marked later thus the theorem is correct.

6

Theorem 2. All the nodes in the network will receive the packet through the termination criteria, when initializing one source node as the forward node. Proof. When one source node is a forward node and is marked, then from theorem 1, the one-hop neighbor of source will be marked later. By iteratively putting theorem1 into application, all the nodes will be marked later. 3.2

Modified Relayed

In the modified relay, node v will not rebroadcast the packet if it checks its status as having received the packet previously, irrespective of whether it was selected as a forward node or not. Theorem 3. If a node v receives the packet from node u and is not selected by u as its forward node, then there is no need for another node w to select v as its forward node. Proof. If node u does not select v as its forward node, then it will select other nodes to cover N (N (u)) including N (v), so N (v) will later be set as relayed, so does not need w to select v as the forward node to cover N (v)∪N (N (w)).

4

Example

0 1 2

3

4

5

8 7 11

9

10 1 4

1 6

13

12

Fig. 4. Example MANET topology with source node 0

Figure 4 shows the connectivity map of the MANET environment used in our example to illustrate the difference between the forwarding algorithms and termination criteria. Table 4 depicts the differences between “modified” and “original”

7 Table 1. Forward nodes list for Figure 4 with different algorithm and termination criteria Original Marked Modified Marked Original Relayed Modified relayed DP 0,1,2,3,4,7,8,10,12 0,1,2,3,7,8,10,12 0,1,2,3,4,7,8, 9,10,12,13 0,1,2,3,7,8,10,12,13 PDP 0,1,2,3,4,7,8,10 0,1,2,3,7,8,10 0,1,2,3,4,7,8,9,10,13 0,1,2,3,7,8,10,13 TDP 0,1,2,3,4,7,8,10 0,1,2,3,7,8,10 0,1,2,3,4,7,8,10 0,1,2,3,7,8,10

in the DP algorithm. Using the original marked termination algorithm, node 2 will select node 4 as its forward node, and for the termination criteria, node 4 will do the rebroadcast, since node 6 is unmarked. When the new modified marked termination algorithm is used, node 2 will not choose node 4 as its forward node. This is because after node 1 has selected node 3 as its forward node, node 4 will be marked, then node 2 will drop node 4 out of consideration as a forward node. In the relay termination algorithm, node 13 will select node 9 as its forward node. Since node 9 has not relayed the packet before, it will act as the forward node. In the case of modified relay, node 13 will select node 9 as its forward node. When node 9 receives the packet, based on the new but termination criteria, node 9 will not rebroadcast the packet since its status is relayed. In this example, we have illustrated how the different dominant nodes selection algorithms have benefited from using the “modified” termination mechanism. The number of forward nodes is reduced in this example.

5

Simulation results

Simulations are done using the unit disk graph [10]. The set-up for the simulation is as follows: – Graph area = 100x100 – Transmission range = 40 – Number of nodes = 20 to 100 A total of 400 sample graphs are generated and for each graph 20 nodes are selected as the broadcast source. The results on number of forward nodes are then averaged across all the experiment and plotted. The experiments were carried out for the three dominant node selection algorithms: DP, TDP, and PDP. Figure 5, Figure 6 and Figure 7 show the average number of forward nodes with different termination criteria and different dominant node algorithm. In all the figures, AMCDS is the lower bound for the performance as it’s a centralized system, while the others use a distributed algorithm. All the algorithms perform poorly when they use the relay termination criteria. This is to be expected as it only makes use of its own information, i.e. whether it had previously forwarded the packet. The other termination criteria makes use of its 1-hop neighbors information.

8

55

mark modified mark relay modified relay AMCDS

50

number of forward node

45 40 35

DP

30 25 20 15 10 5 0 20

30

40

50

60

node number

70

80

90

100

Fig. 5. Average number of forward nodes for DP algorithm

20

mark modified mark relay modified relay AMCDS

18

number of forward node

16 14

TDP

12 10 8 6 4 2 0 20

30

40

50

60

node number

70

80

90

Fig. 6. Average number of forward nodes for TDP algorithm

100

9 24

mark modified mark relay modified relay AMCDS

22

number of forward node

20 18

PDP

16 14 12 10 8 6 4 2 0 20

30

40

50

60

node number

70

80

90

100

Fig. 7. Average number of forward nodes for PDP algorithm

The modified termination criteria reduces the number of forwarding nodes. The performance of the algorithm using different termination criteria increases as the number of nodes increase. Example for node number of 100, the percentage reduction in the number of forward nodes based on modified marked compared with original marked termination criteria is 16.8% for DP, 14.6% for PDP, and 11.7% for TDP. The percentage of the reduced forward nodes based on modified relay compared with the original relay is 78.3% for DP, 60.0% for PDP and 51.5% for TDP.

6

Conclusions

In this paper, we have summarized some previously promising algorithms to select the forward nodes, and related termination criteria to determine whether node should broadcast or not. Since the number of forward nodes depends largely on the termination criteria, we have modified one termination criteria and proposed a hybrid one. From the simulation results, we can see that the new termination mechanisms show marked improvement when incorporated with existing forwarding node selection algorithms, especially when the number of nodes in the network increases.

References 1. C. Perkins, E. Royer, and S. Das, ”Ad hoc on demand distance vector (AODV) routing”, Internet Draft: draft-ietf-manet-aodv-09.txt, Nov. 2001.

10 2. D. Johnson and D. Maltz, ”Dynamic source routing in ad hoc wireless networks”. In T. Imelinsky and H. Korth, editors, Mobile Computing, Kluwer Academic publishers, pp.153-181, 1996. 3. Y.Ko and N.H. Vaidya, ”location-aided routing (LAR) in mobile ad hoc networks”, In proceedings of the ACM/IEEE International Conference on Mobile Computing and Networking (MOBICOM), pp.66-75, 1998 4. Zygmunt J. Haas, Marc R. Pearlm, and prince Samar, ”The Zone Routing Protocol (ZRP) for ad hoc networks”, IETF MANET Internet Draft, July 2002. 5. B. Das, R. Sivakumar, and V. Bharghavan, ”Routing in Ad hoc Networks Using a Virtual Backbone”, Proc. Int’l Conf. Computer Comm. And Networks ’97, pp. 1-20, Sept. 1997. 6. A. Qayyum, L. Viennot, and A. Laouiti, ”Multipoint Relaying for Flooding Broadcast Message in Moble Wireless Networks”, Proc. 35th Annual Hawaii International Conference on System Science-2002. 7. J. Wu and W. Lou, ”Forward-Node-Set-Based Broadcast in Clustered Mobile Ad hoc Networks”, Technical Report CSE-02-15, June 2002. 8. H. Lim and C. Kim, ”Flooding in Wireless Ad hoc Networks”, computer Comm. J.,vol. 24, no. 3-4, pp. 353-363, 2001. 9. W. Lou and J. Wu, ” On Reducing Broadcast Redundancy in Ad hoc Wireless Ad hoc Networks”, IEEE Transactions on Mobile Computing, vol. 1, no. 2 APRILJUNE 2002. 10. B.N. Clark, C.J.Colbourn, and D.S.Johnson, ”Unit Disk Graphs”, Discrete Math., vol. 86, pp. 165-177, 1990. 11. S. Ni, Y. Tseng, Y. Chen, and J. Sheu, ”The Broadcast Storm Problem in a mobile Ad hoc Networks”, Proc.MOBICOM ’99, pp. 151-162, Aug. 1999.