Hop Distance Based Routing Protocol for MANET

Transcription

1 Hop Distance Based Routing Protocol for MANET Filip Cuckov, Min Song Department of Electrical and Computer Engineering Old Dominion University 231 Kaufman Hall Norfolk, VA Abstract The Hop Distance based Routing Protocol (HDR) is an on-demand based routing protocol suited for mobile ad hoc networks integrated with wired networks. HDR considers an efficient way to build multi-hop routes with low routing overhead, through efficient route discovery and route maintenance mechanisms. HDR does this by utilizing the hop count information between mobile nodes and access points to localize the route discovery and maintenance within a limited area in order to reduce routing overhead. HDR is an adaptive protocol and does not require any user or network dependent predefined inputs in order to operate. Simulation results show that HDR is highly efficient in reducing node participation and overhead in route request operations. Keywords- hop count; MANET; routing protocol I. INTRODUCTION Most current on-demand routing protocols applicable to mobile ad hoc networks (MANETs) require network-wide flooding to discover the location of the destination node and to discover the shortest route to it. Network flooding when deduced to node-level broadcasting proves to be a fundamental operation that serves many purposes, among which could be neighbor discovery, collection of global information, naming, addressing, etc [1]. Network wide flooding reduces the network capacity by sending information to mobile hosts which are not supposed to receive it, thus increasing the traffic load, which also increases individual node power consumption. Even though flooding may be a simple solution to many networking problems, it is expensive in terms of communication overhead as it causes redundancy, contention and possibly collision. Therefore, in order to increase the efficiency within a MANET, it is of great interest to reduce or avoid flooding in all communications. To alleviate network flooding, broadcasts and rebroadcasts of packets should be eliminated or limited. Several existing approaches attempt at solving this problem, some dependent on global positioning data, some on an input parameter or threshold mechanism to help a mobile host to decide whether to employ unicasting or broadcasting. The problem with these approaches is that they depend on either special information (like GPS coordinates) or a provided user input that is adjusted for a given network. In cases where the topology and density of a MANET changes dynamically and rapidly, these approaches may either consume massive computing power to operate or may not provide the expected reduction in network flooding. Adaptive approaches and algorithms for MANETs have shown to be a more robust and reliable in combating the flooding problem in most communications The issue of primary interest in this paper is how to reduce the initial flooding of a route request (RREQ) process in ondemand routing. Regardless of any specific implementation of on-demand routing protocols [6, 7, 8], the route discovery procedure occurs when a source node desires a route to a given destination node. Typically, the source node sends a RREQ packet to its immediate neighbors, which re-broadcast the same message to their immediate neighbors and so forth, resulting in a network-wide flooding until the destination node has received the message and sends a route reply (RREP) packet to the source node via the minimal cost (could be minimal hop count, delay, etc) reverse route. The contribution of this paper is a novel Hop Distance based Routing (HDR) protocol that utilizes hop count information and reduces the area of flooding by only including a small subset of the nodes in the network for the RREQ process. Some possible scenarios, where HDR s employment could provide significant benefit, are in access point (AP) based infrastructures such as military combat applications, telemedicine, crisis management and mobile educational systems. In scenarios such as these mobile nodes are connected to a command center and are in need to transmit real-time information efficiently between themselves and to the command centers for control purposes. The rest of the paper is organized as follows. Section II reviews the related work. Section III presents the hop distance based routing protocol for MANETs. The protocol is verified in Section IV via simulations. Finally, Section V provides the conclusions. II. RELATED WORK Several approaches have been defined to reduce network flooding in RREQ processes [1, 2], which are based on either avoiding redundant broadcasts, or utilizing location-based information to define a smaller area of flooding. HDR falls into the second category as it uses hop counts to minimize the RREQ region within the network. The first type of approach is based on employing techniques to identify broadcasts that are redundant and have expired in terms of usability. The basic idea behind this approach is that the more a node hears the same broadcast packet; each consequent re-broadcast of the same packet on its part will result in low or no additional coverage of the network. This concept is straightforward, because during the first /7/$ IEEE 11

2 transmission of the packet the node will cover all its neighbors, and on each following re-transmission the same nodes will be covered and possibly nodes that have moved within the transmission range of the forwarding node. Naturally, the way to minimize re-broadcasts is to employ a counter-based scheme (per packet) and limit the number of transmissions. A certain threshold in these counter-based schemes has to be defined, depending on the size and density of the given network, in order for these schemes to provide a benefit. The location-based approach has many implementations, as a location of a node could be defined in either concrete (as in GPS coordinates) or abstract terms. In a case where each node within a network is equipped with a positioning device, the problem of reducing or eliminating redundant communication becomes much easier, since accurate positioning information could be used to calculate the coverage of each node and routing could be based on exact geographical directions [8]. In the case where the positioning of nodes in the network is defined by hops, measured relative to other nodes or an AP, the approaches to reduce flooding are mostly based on simple accounting [6]. By employing a counter in each packet, a given node can calculate whether it should participate in the ongoing communication. A common approach is an expanding ring search (ERS), where the source (S) sends a RREQ in a radius of x hop counts, and if it fails to obtain a route to the destination (D) it increases the hop count radius until D is reached. This approach, if not employed properly with a predefined threshold of the rate of expansion of the hop count ring, could prove to be worse than simply flooding the network. A Query Scope Agent (QSA) is defined in [5], which assists in the selection of an appropriate ERS approach that will be effective in reducing RREQ communication overhead. QSA takes in as input a maximum allowable value for the route discovery radius and estimates network parameter values before selecting the threshold for a given ERS. The network parameter values include the number of pairs of communicating nodes, a count of k-hop neighborhoods, ERS type success probabilities and hop delay. Following a probabilistic model, QSA adjusts a threshold parameter for ERS that is predicted to give the best performance in reduction of RREQ communication overhead for a given network. QSA, effectively an adaptive sub-flooding mechanism, proves to be an efficient method of utilizing hop counts to reduce routing overhead. III. THE HOP DISTANCE BASED ROUTING PROTOCOL A. Basic Operation HDR is a an on-demand routing protocol that utilizes hop count information to reduce the number of nodes participating in a RREQ process and obtains the lowest hop count path to a given destination. The protocol requires each node in the network to obtain a hop count and to register with the AP, after which the given network is divided in hop count regions, where the AP contains the logical view of the topology. Since Figure 1. HDR node operation state diagram. all nodes operate as routers, the can participate in RREQs, RREPs or data forwarding operations, as depicted in Figure 1. Given that a source node initiates a RREQ, each node within the network that receives this request can determine whether it falls in the critical region for the route discovery as defined by HDR and whether it should participate in the given operation. In terms of complexity, HDR is a protocol that is fairly simple to implement and one that uses low computing resources for operation. The bulk of the computing overhead in an HDR node is in memory operations, since each node has to contain a collection of recently heard messages in order to avoid rebroadcasts of the same, yielding a computational complexity of O(n) for the manipulation of that collection. The processing complexity of HDR is constant, since the most complex operation is the comparison of a node s hop count to a given value. Regarding the networking complexity, HDR exhibits a linear, O(n), communication overhead, as it can be seen from the simulation results in section IV. HDR operates in the following fashion. First, each node obtains its hop count (HC) from an AP advertisement message, which contains a hop counter that is incremented as it hops between nodes. The AP emits advertisement messages at intervals of T seconds, a value adjusted based on the dynamics of the nodes within the network. Each node records the lowest hop count it encounters from these advertisement messages. Then logically, the network is divided in hop count regions as shown in Figure 2a. The shaded square in Figure 2 represents the AP and the circles represent the nodes in the network. The concentric circles centered at the AP represent hop count regions in the network, meaning that the nodes closest to the AP have a hop count of one and so forth. In actuality the shape of the hop count regions of any given network will be defined by the transmission range of the individual nodes and will not necessarily resemble the ones as shown in Figure 2, but for the sake of simplicity and the examples that follow Figure 2 s depiction will suffice. The task of HDR is to minimize the route discovery region by involving a smaller subset of nodes within the network. Given a sample scenario, where S has a hop count of two and D has a hop count of one, HDR would minimize the subset of nodes that participate in the route discovery phase to the nodes represented as shaded in Figure 2b. 12

3 B. HDR Implementation Details The RREQ and RREP event handlers, presented in pseudocode below, represent the route discovery mechanism of the HDR protocol. Figure 2. (a) Logical network representation after AP advertisement package (b) Nodes involved in HDR route discovery. The operation of HDR is described following the example given in Figure 3. When S desires a route to D, it initiates a route discovery by broadcasting a route-request packet (RREQ). The route discovery area is defined by utilizing each node s hop count that is based on the distance to the AP. Figure 3 depicts an intersection of two circles with radii r1 and r2, respectively. It can be seen that the nodes within the intersection of the logical circles are the ones which will participate in the RREQ from S to D. The radii are drawn in terms of hop count distance, rather than actual distance. The value of r1, centered at the AP, is the maximum hop distance of the S and D hop counts. This area encompasses all nodes which, in this specific case, have a hop count value of two or less. This guarantees that both the S and D are covered in this area and that in the worst case all the nodes within this area might have to participate in the route discovery phase. The value of r2, centered at S, is the sum of the hop counts of the S and D. Again this defines an area where D would definitely fall in. HDR recognizes that nodes that have a larger hop count than either the source or the destination should not be included in the route discovery procedure. HDR also recognizes that if the hop count of the source node is x and the hop count of the destination node is y, than any nodes that are x+y hops away from the source should also not be included in the route discovery routine. The intersection of the two regions is the desired route discovery region of HDR, which is the basics of how the protocol minimizes unnecessary communication, by involving a smaller subset of the nodes in the route discovery phase. Figure 3. Critical area defining participating nodes in route discovery. receive RREQ (packet) { //Ignore the RREQ if RREQ originated from this node If (Sid = CurrentNodeID) { return } received = InspectRecentMessages(packet) if (received) { return } if (Did = CurrentNodeID) //node is the destination { Route Route (+) CurrentNodeID //route appended send RREP SaveRoute //save the route to S return } if ( CurrentNodeHC > max (Shc, Dhc) ) { return } // r1 check if ( Is > Shc + Dhc) { return } // r2 check //if message passed all the above tests then it is a unique RREQ SaveUniqueMessage Route Route (+) CurrentNodeID //route appended send RREQ //broadcast to other nodes } receive RREP (packet) { if( NEXTid = CurrentNodeID) { if (Destination = CurrentNodeID) if ( Is for route is less than existing one or no record exists) SaveRoute else NEXTid = next node on route to S send RREP (packet) } } The RREQ event handler processes the S hop count (Shc) the D hop count (Dhc) and a hop count counter from the S to an intermediate node I called Is, which increments by one for each hop the message takes. When S sends a RREQ message all nodes that have a hop count of more than the maximum of the hop counts of S and D will disregard the message. If an intermediate node has a hop count value that satisfies the previous condition, then it looks at the Is value and if it is greater than Shc + Dhc the node will also disregard the message. This means that only nodes in the intersection region defined in Figure 3 will participate in the route discovery phase. Also each node will disregard any message that has originated from itself, or if it has been already received. If the intermediate node receiving the RREQ message is in the RREQ region, then it will add its own ID to the Route field in the message, increment Is, and will re-broadcast the message to its neighbors. If the node is the destination of the message, then the route to S is saved and a RREP is returned via the exact route the message has traveled to D. The RREP event handler replies the path to the source if the current node is the destination; otherwise it forwards the message to the intended recipient. Any intermediate node will disregard an incoming RREP if it is not the next recipient of the message, otherwise it will set the next hop recipient to be the next node in the Route field and will forward the message to the intended next recipient. Once the destination of the RREP (or the S of the RREQ) receives the message it will save 13

4 the route to D if there is no current route or if the freshly received RREP contains a route that has less number of hops than the existing route on record. This way HDR guarantees that the path with the minimal number of hops is selected for data forwarding purposes. IV. SIMULATION AND RESULTS Simulations of HDR operation, developed in the C language, were conducted on multiple scenarios to examine the protocol s effectiveness in reducing flooding. The results were based on a simulation setup where a pre-defined number of nodes were placed randomly, based on a uniform probability distribution, in a field of size 1 m by 1 m. Each node had a transmission radius of 12.5 m, as well as the AP which was placed in the middle of the field. In order to obtain an unbiased average of a measure of interest, the simulations were run for one thousand RREQs originating from a random S to a random D. The results obtained from the simulations are displayed in Figures 4-8. These results demonstrate HDR s effectiveness in efficient route discovery and low overhead during the RREQ process. It is important to mention that the curves report the average of a large number of simulated RREQs over differing network densities, meaning that even though HDR may provide better performance in some scenarios, the expected performance is portrayed by these curves. Figure 4 describes the performance of HDR by illustrating the average percentage of nodes participating in the RREQ process over 1 RREQs, as a function of network density. Namely, as the number of nodes in the network increases, Figure 4 illustrates that the percentage of nodes that participate in the RREQ process increases too. This result is expected, since the denser the network, the more redundant communication there will be. For 5 nodes within the network, the percent of actively participating nodes in the HDR protocol is 26%, while as the density increases steadily to 1 nodes the percentage of participating nodes also increases steadily to 35%. In the range from 1 to 15 nodes in the network, the percentage stays at a steady level and it increases to 43% once the network contains 2 nodes. RREQ Network Utilization Figure 4. Average percentage of nodes participating in RREQs. Avg. RREQ Message Overhead Figure 5. Average network message overhead per RREQ. The steadiness in level, also observed in the range from 2 to 25 nodes, is due to the fact that increasing the density by a given amount has little or no impact on the interconnectivity and thus added redundant communication within the network. In fact, as the simulation results showed, in the range from near 1 up to 175 nodes, the network contains sufficient number of nodes, is well connected and displays an average of about 35% node participation in the RREQ process by utilizing the HDR protocol. As the density of the network increases to excessive levels, HDR s performance decreases and an upper limit of 5% node participation is registered. From Figure 4 it could be concluded that in a well connected network HDR is expected to employ on average 35% of the nodes, while as the density of the network increases HDR s performance is expected to decay gracefully to a level where up to 5% of the nodes are used for the same process. Figures 5 and 6 describe the packet overhead per RREQ in the network as a function of network density and the same overhead per node, respectively. Figure 5 suggests a linear increase in routing overhead as the density of the network increases, which is a respectable property of HDR. This result was expected, as according to Figure 3 as network density increases, the number of nodes and the amount of RREQ communication in the critical intersection region should increase too. In order to better understand HDR s impact on individual nodes, Figure 6 illustrates the same information in terms of an average number of packets processed per RREQ as a function of network density. From this figure it can be seen that as the density in the network increases, the average packet overhead per node increases in a logarithmic manner. In a well connected scenario, where the number of nodes in the network ranges from near 1 up to 175 nodes, nodes utilizing HDR are expected to process little more than one packet for any two RREQs. Figure 7 depicts the average RREQ overhead in terms of node utilization as the number of nodes within the network increases. From this figure it can be seen that QSA at its worst reported performance, labeled as QSA LT (low threshold), has a linear increase in routing overhead, similarly to HDR, while at its best, labeled as QSA HT (high threshold), the overhead 14

5 Routing Overhead per Node Figure 6. Average number of RREQ messages per node vs. network density. Avg. RREQ Message Overhead QSA LT QSA HT HDR Number of nodes Figure 7. Average RREQ Overhead in QSA and HDR. tends to drop of rapidly as the network density increases to the point where it matches the performance of HDR. Figure 8 displays the average percentage of nodes participating in the RREQ process as a function of network density, for QSA s high and low threshold and HDR accordingly. The network density ranges from a sparse to a well connected network and QSA in this range displays at its best an average of around 5% node utilization for RREQs. HDR in this range displays a steadily increasing percentage of node utilization that averages at around 35%. However, it seems that as network density increases QSA seems to be more effective at reducing the percentage of RREQ participating nodes, as HDR shows an increasing trend in the same range. From these data it could be concluded that in a well connected network with a reasonable node density HDR outperforms QSA in reducing both routing overhead and minimizing the flood region for the RREQ process, while QSA with high threshold shows promising results in highly dense networks. The simulation results, illustrated in these figures, show that HDR is an effective protocol that achieves its aim of introducing low routing overhead, through efficient route discovery and route maintenance mechanisms during the RREQ process. RREQ Network Utilization QSA LT QSA HT HDR Number of nodes Figure 8. Percent of Network utilized for RREQ in QSA and HDR. V. CONCLUSIONS This paper presented a novel Hop Distance based Routing protocol for MANETs that effectively reduces network flooding during the RREQ process, using only hop count information from nodes. Through simulations it was observed that the overhead introduced by HDR is low relative to other approaches. HDR exhibits low per node message overhead, and proves to be highly effective, only incorporating on average 35% of all nodes for any given RREQ process, in networks whose interconnectivity ranges from low to relatively high. In extremely dense networks where there is high redundancy in communication, HDR provided results that match the expected performance. REFERENCES [1] Yu-CheeTseng, Sze-Yao Ni, and En-Yu Shih. Adaptive approaches to relieving broadcast storms in a wireless multihop mobile ad hoc network. 21st International Conference on Distributed Computing Systems. April 21, pp [2] Peng, W. and Lu, X. On the reduction of broadcast redundancy in mobile ad hoc networks. Proc. of the 1st ACM international Symposium on Mobile Ad Hoc Networking & Computing. 2, pp [3] Pucha, H., Das, S. M., and Hu, Y. C. The performance impact of traffic patterns on routing protocols in mobile ad hoc networks. Proc. of the 7th ACM international Symposium on Modeling, Analysis and Simulation of Wireless and Mobile Systems. October, 24, pp [4] Das, S.R., Perkins, C.E., Royer, E.M. Performance comparison of two on-demand routing protocols for adhoc networks. INFOCOM 2. pp [5] Sucec, J., Marsic, I. A Query Scope Agent for Flood Search Routing Protocols. Wireless Networks 23. Vol. 9, No. 6, pp [6] Kuruvila, J., Nayak, A., Stojmenovic, I. Hop count optimal positionbased packet routing algorithms for ad hoc wireless networks with a realistic physical Layer. IEEE Journal on Selected Areas in Communications. June 25, Vol. 23, Issue 6, pp [7] Basagni, S., Chlamtac, I., Syrotiuk, V. R., and Woodward, B. A. A distance routing effect algorithm for mobility (DREAM). Proceedings of the 4th Annual ACM/IEEE international Conference on Mobile Computing and Networking. October [8] Ko, Y.-B. Vaidya, N. H. Location-Aided Routing (LAR) in mobile ad hoc networks. Wireless Networks 2. Vol. 6; Part 4, pp [9] D. Obradovic. Formal Analysis of Routing Protocols. PhD thesis, University of Pennsylvania,