Efficient Location-Aided Flooding Scheme Initiated by Receiver for MANETs* Ngoc Duy Pham and Hyunseung Choo

†

School of Information and Computer Engineering Sungkyunkwan University, Korea

[email protected], [email protected] ABSTRACT Flooding is a fundamental, critical, and indispensable operation for supporting various applications and protocols in wireless ad hoc networks. The traditional flooding scheme generates excessive redundant packet retransmissions, however, causing contention and packet collisions, and ultimately wasting precious limited bandwidth and energy. Recently, some flooding schemes have been studied to avoid those problems, but these algorithms either perform well in redundant transmissions or require that the node maintain information about neighbors more than one hop away. One of the most efficient approaches is found in the work of Liu et al., which uses only 1-hop neighbor information. The advantage of this scheme is that it achieves local optimality in terms of the number of retransmitting nodes, although it still produces many redundant transmissions. In this paper, we propose an efficient flooding protocol that minimizes flooding traffic by leveraging location information of 1-hop neighbor nodes. Our scheme is receiver-based, which means that each receiver of a flooding message determines whether it should forward the message based on the given retransmission rule. Simulation results show that our scheme is highly efficient. It is able to reduce the number of forward nodes almost to that of the lower bound but maintains a high delivery ratio.

Categories and Subject Descriptors C.2.2 [Network Protocols]: Routing protocols.

General Terms Algorithms, Performance, Reliability, Design.

Keywords Broadcasting, Receiver-based, Mobile Ad-hoc Networks.

1. INTRODUCTION Flooding is a simple broadcast protocol for delivering a message to all nodes in a network. Many routing protocols for wireless ad hoc networks [14], such as AODV [4], DSR [10], ZRP [8], and Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. ICUIMC-09, January 15-16, 2009, Suwon, S. Korea. Copyright 2009 ACM 978-1-60558-405-8109101...$5.00.

LAR [11], among others, use flooding as the basic mechanism to disseminate route discovery or to propagate control messages. An efficient flooding scheme is therefore useful for reducing the overhead of routing protocols, decreasing collisions, and improving network throughput. The simplest flooding technique, called pure flooding or blind flooding, was first discussed in [3,9]. In this scheme, every node in the network retransmits the flooding message when it receives a message for the first time. Despite its simplicity and guarantee that a flooding message can reach all nodes (if there are no collisions and the network is connected), pure flooding generates an excessive amount of redundant network traffic, because it requires every node to retransmit the message. Due to the broadcast nature of radio trans-missions, when all nodes flood the message in the network, there is a very high probability of signal collisions, which may cause some nodes fail to receive the flooding message. This is the so-called broadcast storm problem [13,16]. To alleviate this problem, several efficient flooding schemes have been presented for mobile wireless ad hoc networks (MANETs). One trend in flooding scheme designs is to apply a connected dominating set (CDS) to flood the control messages [5,17]. CDSbased flooding requires each node to collect 2-hop neighbors' information; maintaining information about more than one hop neighbor is unreliable due to node mobility, however, and this technique results in additional overhead. In Edge Forwarding [18], each node acquires the geographic location information of 1hop neighbors to decide whether it should relay the flooding message. Another noteworthy flooding scheme that uses only 1hop neighbor information is described in the work of Liu et al. [7]. According to this scheme, at each hop, only a subset of neighbor nodes is selected to rebroadcast the flooding message. The simulation results show that the performance of these schemes is considerably better than that of pure flooding; the total number of retransmissions is often much higher than the minimum number of retransmissions required to cover the entire region - the lower bound [15]. 1 In this paper, we propose an efficient flooding protocol that uses only the location information of 1-hop neighbor nodes. In our approach, the receiver is responsible for deciding whether it should forward a message or not. When a node receives a flooding message from a neighbor, it checks whether its own neighbors that

* This research was supported by MKE, Korea under ITRC IITA2008-(C1090-0801-0046). † Dr. Choo is the corresponding author.

-194-

did not receive the message are covered by neighbors that already received the message. If this condition is satisfied, the node should not retransmit the message. Our scheme uses this idea to minimize the number of forwarding nodes and collisions in the network. Performance evaluation shows that our proposed scheme outperforms CDS-based Flooding, Edge Forwarding, and 1HI (Liu's work). The rest of this paper is organized as follows. Related work is discussed in the next Section. Section 3 describes our novel, efficient flooding scheme for wireless ad hoc networks. In Section 4, we compare the performance of various flooding protocols through simulations. Finally, we conclude our work with discussions of future directions for research in Section 5.

set") and attaches this list to the message header. After receiving a flooding message for the first time, all receiver nodes verify whether they are in the sender's forwarding set or not. Every node that is a retransmission node computes its own forwarding set exactly in the same manner that the source node does. Then, based on the geographic location information, the receiver node optimizes its forwarding set by removing the nodes covered by the sender node and other lower-ID, neighboring retransmission nodes. After that, it relays the flooding message with the forwarding set attached. In this manner, the message eventually reaches all nodes. D A B

2. RELATEDWORK

b C

In pure flooding, every node in the network retransmits the flooding message after receiving it for the first time. This flooding mechanism is inefficient, because it leads to serious problems including redundancies and collisions; it is thus not recommended for future works [19]. In the following part, we introduce some efficient flooding schemes that collect extra information and process a message, to reduce the overhead of flooding.

a E

Figure 1: Example of Edge Forwarding.

CDS-based Scheme. A dominating set (DS) is a subset of nodes such that every node in the network is either in the set or adjacent to a node in the set. A CDS is a connected DS, and any routing in MANETs can be achieved efficiently via the CDS [5,15]. In CDSbased schemes, a node marks itself as belonging to the CDS if it has two unconnected neighbors. Therefore, each node must know the network topology of its 2-hop neighbors. One method for obtaining the 2-hop neighbor information is to instruct each node to attach the list of its neighbors to the HELLO message for exchange purposes. This method increases network traffic, however, and causes uncertainty in the 2-hop information due to node mobility. In conclusion, CDS-based flooding, which is receiver-based, is unsuitable for flooding operations under highmobility conditions. Edge Forwarding. Another popular flooding scheme that again uses only 1-hop neighbor information is Edge Forwarding [18]. It is a receiver-based flooding scheme. Apart from maintaining 1hop neighbor information from the HELLO messages in the MAC layer protocols, each node divides its transmission coverage into six equally sized sectors. Upon receiving a flooding message, a node decides whether to forward the control message based on the availability of other retransmission nodes in the overlapped areas. For example, node b in Fig. 1, which has received the flooding message from node a, does not retransmit the message if nodes exist in enclosed zones A, B, and C, since the coverage area of node b is entirely covered by node a and the nodes in zones A, B, and C. In this manner, the number of retransmission nodes and the number of collisions in networks are reduced. The use of only 1hop neighbor information is insufficient, however, to minimize effectively the number of retransmission nodes using this method. Efficient Flooding Scheme using 1-hop Information in Mobile Ad Hoc Networks. A related work, which has the same objective as ours, is that described by Liu et al. [7]. The mechanism of 1HI is summarized as follows: When a source node has a message to be flooded, based on the 1-hop neighbor information, it selects the next-hop retransmission nodes (together called the "forwarding

Regular nodes: a, b Enclosed zones: A, B, C, D, E

Source

Source Retransmission nodes Receiving nodes

Figure 2: Distribution of retransmission of nodes in 1HI. Thus far, the advantage of 1HI is that it is a sender-based flooding scheme that uses only 1-hop neighbor information; the protocol is easy to implement and has a small overhead. This scheme, however, has some disadvantages. First, the forwarding set is only locally optimized based on 1-hop neighbor information; therefore, the number of retransmission nodes is relatively high. Second, the retransmission nodes are densely distributed along the network border (Fig. 2). Most retransmissions broadcasted from networkborder nodes are redundant, because their neighbors have already received this message from other previous-hop retransmission nodes.

3. GROUP FORWARDING SCHEME BASED ON 1-HOP NEIGHBOR LOCATION INFORMATION 3.1 Overview of Method Existing, efficient flooding schemes can be classified into three categories based on the information each node keeps: 1) no need of neighbor information, 2) 1-hop neighbor information, or 3) 2hop neighbor information and more [7]. Our proposed scheme is

-195-

in the second category. One-hop neighbor information can be obtained by exchanging the HELLO message in MAC layer protocols. There are two strategies for choosing forwarding nodes: sender-based, where each sender nominates a subset of its neighbors to be the next hop forwarding nodes, and receiverbased, where each receiver of a flooding message makes its own decision whether it should forward the message. Our approach is also receiver-based.

Suppose r is a node that receives a flooding message for the first time (s is the sender of the message). Using 1-hop neighbor information, r divides its neighbors into three groups (Fig. 3):

We assume that all nodes in the network have the same transmission range R; we also assume that the network is connected. Each node u has a unique ID. Neighbors of u are nodes within the transmission range of u that can receive signals transmitted by u. Node u needs to know the information of its direct neighbors, including their IDs and their geographic locations. The geographic location information can be obtained using location awareness supported by GPS [6], received signal strength indicator [2], time different of arrival [1], and angle of arrival [12].

•

Group#1: Nodes in the transmission range of s whose distance to s is shorter than that from r to s {u | u ∈ c(r ) ∩ u ∈ c(s) ∩ d (u, s) ≤ d (r , s)}

•

Group#2: Nodes in the transmission range of s whose distance to s is larger than the distance from r to s {u | u ∈ c(r ) ∩ u ∈ c( s) ∩ d (u, s) > d (r , s)}

•

Group#3: Nodes that are not in transmission range of s {u | u ∈ c(r ) ∩ u ∉ c(s)}

s

r

Nodes in Group #1 Nodes in Group #2 Nodes in Group #3

3.2 The Proposed Scheme The basic idea of our flooding scheme is as follows: When a node has a message to be flooded to the entire network, it first broadcasts the message to its own neighbors. Once it receives a message from the sender, the receiver divides the nodes inside its coverage area into three groups based on the locations of its neighbors. If the retransmission rule introduced in next part is not satisfied, then the node retransmits the message. Otherwise, the receiver stops forwarding the flooding message. After that, every neighbor nodes receives the message and repeats the same process; upon receiving a flooding message, if the message has been received before, it is discarded; otherwise, the message is delivered to the application layer and the receiver checks whether the rule is satisfied. If not, it transmits the message out.

s Sender (s) r Receiver (r)

Figure 3: Groups division at node r. The following is the pseudo code for dividing the nodes into groups.

Algorithm 1. Grouping neighbor nodes 1 2 3 4 5

Foreach node u in coverage area of r If d(u,s) ≤ d(r,s) then Add u to Group#1 If d(u,s) > d(r,s) then Add u to Group#2 If u ∉ c(s) then Add u to Group#3 Endfor

3.2.1 Group Division

3.2.2 Relaying Message

We aim at designing a 1-hop flooding scheme. We introduce the following definitions:

After dividing neighboring nodes into groups, node r checks the rule to decide whether it should retransmit the message or not.

Definition 1 (Coverage disk of a node). The coverage disk of node u, denoted by c(u), is a disk that is centered at u and whose radius is the transmission range of u.

Retransmission rule. If all nodes in Group#3 are covered by nodes in Group#2, then the retransmission of r is redundant and can be omitted.

Since all neighbors of node u should be covered by c(u), in this paper, we say that "u covers v" or "v is covered by u" when v is neighbor of u.

∀ v ∉ Group#3, v ∉ C(Group#2): r should not retransmit

Definition 2 (Coverage area of a node-set). The coverage area of a set of nodes A, denoted by C(A), is the union of coverage disks of nodes in A. We simply state "the area is covered by A" if the area is within C(A). Definition 3 (Distance between two nodes). The distance between two nodes u and v, denoted by d(u,v), is the Euclidean distance between the positions of those two nodes. The Euclidean distance between two points is calculated as follows:

d (u , v ) = (u x − v x ) 2 + (u y − v y ) 2

Nodes in Group#3 are already inside the coverage area of r; it is clear that when all nodes in Group#3 are covered by nodes in Group#2, then the retransmission by r and nodes in Group#2 cause nodes in Group#3 to receive the same flooding message twice. We can eliminate this redundancy by letting either r or the nodes in Group#2 retransmit the message (but not both). In our proposed scheme, we choose nodes in Group#2 to be responsible for retransmitting the message, because the nodes in Groups#2 cover an area at least as large as the coverage area of r. So, if nodes in Group#2 send messages, then more nodes can receive the flooding message. Besides, the nodes in Group#2 are also further from the sender s than from r, so they can cover a larger area. We give a simple Θ(n 2 ) algorithm to determine the coverage of nodes in Group#3 by the nodes in Group#2. The following pseudo-code determines whether a node should relay the flooding

-196-

1HI

Edge Forwarding

CDS-based Flooding

3000

Lower Bound

Pure Flooding

100

2800 95

2600 2400

deliverability ratio (%)

the number of collisions

ratio of forwarding nodes

Proposed Scheme 1.00 0.95 0.90 0.85 0.80 0.75 0.70 0.65 0.60 0.55 0.50 0.45 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00

2200 2000 1800 1600 1400 1200 1000 800 600 400

0 400

600

800

1000

85 80 75 70 65 60 55

200

200

90

50 200

400

the number of nodes

600

800

1000

200

the number of nodes

400

600

800

1000

the number of nodes

Figure 4: Performance on different number of nodes. Proposed Scheme

1HI

Edge Forwarding

CDS-based Flooding

Lower Bound

Pure Flooding

100 600

1.0

98

0.6

0.4

0.2

94

deliverability ratio (%)

the number of collisions

ratio of forwarding nodes

96 0.8

500

400

300

200

100

92 90 88 86 84 82 80 78 76 74 72

0.0

0 100

150

200

250

300

transmission range

70 100

150

200

250

300

100

transmission range

150

200

250

300

transmission range

Figure 5: Performance on different transmission range. message. In brief, if node r receives the message from s, it divides its neighbors into three groups and then checks whether the retransmission rule is satisfied. Finally, if the rule is not satisfied, the node should retransmit the flooding message to its neighbors.

Algorithm 2. Decision of retransmission 1 2 3 4 5 6 7 8

Foreach node u in Group #3 cover = false Foreach node v in Group #2 If u ∈ c(v) then cover = true break If !cover then return false return true Table 1. Simulation parameters. Parameter

Value

Simulator

ns-2

MAC Layer

IEEE 802.11

Data Packet Size

256 bytes

Bandwidth

2Mb/s

Network Load

10Pkt/s

Number of Nodes

200~1000

Size of Square Area

1000 x 1000m2

4. PERFORMANCE EVALUATION 4.1 Simulation environment We choose the following flooding schemes for comparison with our own: Pure flooding, Edge Forwarding, CDS-based flooding, and the scheme of Liu et al. (1HI). Edge Forwarding is picked because its approach is similar to ours: they divide the coverage area into regions and then check for neighbor nodes within them. CDS is one of the most important techniques for flooding operation in MANETs. We choose [15] as the lower bound to compare with the result of our proposed algorithm. Also, 1HI is chosen as the main target for comparison, because it performs the best among the 1-hop flooding schemes [7]. We run simulations under the ns-2 testbed with the CMU wireless extension. The simulator parameters are listed in Table 1. We use the broadcast mode with no RTS/CTS/ACK mechanisms for all message transmissions. The bandwidth of a wireless channel is set to 2Mb/s as the default. Notice that some of the schemes require the node to send a HELLO message to its 1-hop neighbors periodically. This cost is ignored in our performance study. We analyze flooding efficiency in terms of the following metrics: 1) Forwarding ratio: the ratio of forward nodes in the flooding operation to the total number of nodes in the network. 2) The number of collisions: the sum of the collisions experienced by each node before it receives the flooding message correctly. 3) Delivery ratio: the ratio of the nodes that received packets to the number of nodes in the network for one flooding operation.

-197-

In each simulation run, we generate a certain number of nodes and randomly place them within a square area. The source that initiates a flooding message is randomly picked from nodes in the network. Only one flooding occurs at any one time (except in the experiments on delivery ratio). Three flooding schemes and the theoretical lower bound that are mentioned above are simulated and compared with our scheme under the same conditions. The results presented in the following figures are the means of 100 separate runs.

4.2 Results and Discussion 4.2.1 Performance on the number of nodes In this simulation, a certain number of nodes, from 200 to 1000, are randomly placed on a 1000 x 1000m2 area. The transmission range is fixed at 250m. In the experiment on delivery ratio, the network load is set to 10Pkt/s, that is, the network generates 10 flooding messages per second on average. The deliverability is calculated over 100 seconds. •

The performance of our scheme is significantly better than the performance of Edge Forwarding, CDS-based, and Liu's schemes as shown in Fig. 4. When the number of nodes in our proposed scheme reaches 1000, only 6% of the total number of nodes participates in forwarding, whereas for Edge Forwarding, CDS-based Flooding, and 1HI, 51.1%, 69.5%, and 16.9% participate in forwarding, respectively.

•

Our scheme and the Edge Forwarding, CDS-based, and Liu's schemes incur many fewer collisions than does pure flooding. The reason is that every node forwards the message in pure flooding, resulting in large number of collisions.

•

As shown in Fig. 4, the deliverability of our scheme is significantly higher than the ratios of Edge Forwarding, CDS-based, and 1HI. The delivery ratio of other methods are only 54–92%, our scheme achieved nearly 100% deliverability when the number of nodes ranges from 600 to 1000 in our experiment (corresponding to dense networks).

4.2.2 Performance on transmission range In this simulation, 500 nodes are randomly placed on a 1000 x 1000m2 area. We study the performance against the transmission range of each node. The simulation results are plotted in Fig. 5. We make the following observations: • •

Performance of our scheme is significantly better than the performance of the target papers' work. Fig. 5 shows that when we adjust the radio range, our proposed scheme guarantees the same high deliverability (close to 100%) and exhibits considerably greater deliverability than the other methods. Besides, it is shown clearly that increasing the transmission range not only causes more collisions, but also provides more chances for nodes to receive the flooding message from other nodes.

5. CONCLUSION Traditional approaches to the implementation of flooding suffer from excessive redundancy of messages, resource contention, and signal collisions. In this paper, we address the efficient flooding problem in a wireless ad hoc network. The paper presents an efficient flooding scheme that uses only 1-hop neighbor information. Simulation results have shown that our proposed scheme uses fewer forwarding nodes, incurs fewer collisions, obtains a higher deliverability, and is more highly scalable than the existing schemes. In future work, we will study how to adapt our scheme to a mobile environment, that is, an environment in which every node moves around. In particular, we will study the important issue of how to check the satisfaction of the retransmission rule under these conditions.

REFERENCES [1] C. H. A. Savvides and M. Srivastava. Dynamic Fine-Grained Localization in Ad hoc Networks of Sensors. In Proceeding of the Annual International Conference on Mobile Computing and Networking, pages 166–179, 2001. [2] P. Bahl and V. Padmanabhan. RADAR: An In-Building RFBased User Location and Tracking System. In Proceeding of the Conference on Computer Communications, pages 775– 784, 2000. [3] G. T. C. Ho, K. Obraczka and K. Viswanath. Flooding for Reliable Multicast in Multi-hop Ad Hoc Networks. In Proceeding of the International Workshop on Discrete Algorithms and Methods for Mobile Computing and Communication, pages 64–71, 1999. [4] E. B.-R. C.E. Perkins and S. Das. Ad Hoc On-Demand Distance Vector (AODV) Routing. Internet experimental rfc 3561, 2003. [5] F. Dai and J. Wu. An extended localized algorithm for connected dominating set formation in ad hoc wireless networks. IEEE Transaction on Parallel and Distributed Systems, 15:908–920, 2004. [6] I. Getting. The global positioning system. IEEE Spectrum, 30:36–47, 1993. [7] P. W. X. L. F. Y. H. Liu, X. Jia. A distributed and efficient flooding scheme using 1-hop information in mobile ad hoc networks. IEEE Transaction on Parallel and Distributed Systems, 18:658–671, 2007. [8] Z. Haas and M. Pearlman. The Zone Routing Protocol (ZRP) for Ad Hoc Networks. Internet draft – mobile ad hoc networking (monet) working group of the internet engineering task force (ietf), 1997. [9] D. M. J. Jetcheva, Y. Hu and D. Johnson. A Simple Protocol for Multicast and Broadcast in Mobile Ad Hoc Networks. Internet draft: draft-ietf-manet-simple-mbcast-01.txt (outdated), 2001. [10] D. Johnson and D. Maltz. Mobile Computing: Dynamic Source Routing in Ad Hoc Wireless Networks. Kluwer Academic Publishers, Dordrecht, The Netherlands, 1996.

-198-

[11] Y. Ko and N. Yaidya. Location-Aided Routing (LAR) in Mobile Ad Hoc Networks. In Proceeding of the Annual International Conference on Mobile Computing and Networking, pages 66–75, 1998.

[16] Y. C. S. Ni, Y. Tseng and J. Sheu. The broadcast storm problem in a mobile ad hoc network. In Proceeding of the Annual International Conference on Mobile Computing and Networking, pages 151–162, 1999.

[12] D. Niculescu and B. Nath. Ad Hoc Positioning System (APS) using AoA. In Proceeding of the Conference on Computer Communications, pages 1734–1743, 2003.

[17] J. Wu and H. Li. On Calculating Connected Dominating Set for Efficient Routing in Ad Hoc Wireless Networks. In Proceeding of the International Workshop on Discrete Algorithms and Methods for Mobile Computing and Communications, pages 7–14, 1999.

[13] R. S. P. Sinha and V. Bharghavan. Enhancing Ad hoc Routing with Dynamic Virtual Infrastructures. In Proceeding of the Annual Joint Conference of the IEEE Computer and Communications Societies, pages 1763–1772, 2001. [14] C. Perkins. Ad Hoc Networking. Addison-Wesley Professional, Boston, 2001. [15] K. A. P.J. Wan and O. Frieder. Distributed Construction of Connected Dominating Set in Wireless Ad hoc Networks. In Proceeding of the Conference on Computer Communications, pages 1597–1604, 2002.

[18] K. H. Y. Cai and A. Phillips. Leveraging 1-hop Neighborhood Knowledge for Efficient Flooding in Wireless Ad Hoc Networks. In Proceeding of the International Performance Computing and Communications Conference, pages 7–9, 2005. [19] S. N. Y. Tseng and E. Shih. Adaptive Approaches to Relieving Broadcast Storms in a Wireless Multihop Mobile Ad hoc Networks. In Proceeding of the International Conference on Distributed Computing Systems, pages 481– 488, 2001.

-199-