2012 26th IEEE International Conference on Advanced Information Networking and Applications

A Node-Disjoint Multipath Routing Method based on AODV protocol for MANETs

Chhagan Lal1 ,V.Laxmi2 , M.S.Gaur3 ,
Department of Computer Engineering Malaviya National Institute of Technology, Jaipur,Rajasthan,India chhagan1 vlaxmi2 gaurms3 @mnit.ac.in on multipath routing protocols for MANETs is presented in [1]. Node-disjoint multipath routing allows the establishment of multiple paths, each consisting of an unique set of nodes between a source and destination. We know that MANETs consist of mobile nodes that cause frequent link failures. This link failure causes two main problems. Firstly, when a route break occurs, all packets that have already been transmitted on that route are dropped and it decreasing the average packet delivery ratio (PDR). Secondly, the transmission of data trafc is halted for the time till a new route is discovered and it increasing the average end-to-end delay. In this paper, we develop a node-disjoint multipath routing method (NDMP-AODV) based on ad-hoc on-demand distance vector (AODV) [2] routing protocol. Our proposed approach minimizes the effect of link failure. Hence, the above mentioned two problems caused by frequent link failures are addressed. NDMP-AODV ensures that after a route is broken, the node can continuously send data without any delay, using one of the backup routes stored in its routing table during route discovery process. It has been shown that NDMP-AODV discovers multiple paths with a very low routing overhead as compared to other existing multipath AODV protocols. The remainder of the paper is structured as follows. In Section II, we present related work in our area by providing a brief description of existing multipath extensions of AODV routing protocol. The proposed method NDMP-AODV used for discovering multiple paths is presented in Section III. In Section IV, we present the experimental setup details and provide results with analysis obtained through various simulations. Finally, the conclusions and directions for future work are provided in Section V. II. R ELATED W ORK In this section, we discuss the previous work done on multipath routing methods based on AODV protocol. Multipath routing creates multiple paths between a source-destination pair. In case of the failure of rst route, the backup routes are used for continues data transmission. In multipath routing protocols, the paths between a source and destination can be link-disjoint, node-disjoint or non-disjoint. In [3], authors present a node-disjoint multipath extension for AODV referred as MP-AODV. MP-AODV discovers two
399

AbstractFrequent link failures are caused in mobile ad-hoc networks due to nodes mobility and use of unreliable wireless channels for data transmission. Due to this, multipath routing protocols become an important research issue. In this paper, we propose and implement a node-disjoint multipath routing method based on AODV protocol. The main goal of the proposed method is to determine all available node-disjoint routes from source to destination with minimum routing control overhead. With the proposed approach, as soon as the rst route for destination is determined, the source starts data transmission. All the other backup routes, if available, are determined concurrently with the data transmission through the rst route. This minimizes the initial delay caused because data transmission is started as soon as rst route is discovered. We also propose three different route maintenance methods. All the proposed route maintenance methods are used with the proposed route discovery process for performance evaluation. The results obtained through various simulations show the effectiveness of our proposed methods in terms of route availability, control overhead, average end-to-end delay and packet delivery ratio. Index Termsmobile ad-hoc routing, Multipath, node-disjoint, Route availability, AODV protocol, Video transmission

I. I NTRODUCTION Mobile ad hoc networks (MANET) are collection of mobile hosts which are self-congurable, self-organizable and selfmaintainable. These mobile hosts communicate with each other through wireless channels with no centralized control. The inherently infrastructure-less, inexpensive and quick-todeploy nature of MANETs is providing a promise for its use in diverse domains. Over the years, multimedia streaming over the internet has established well with numerous applications including audio/video streaming, TV on demand, voice-overIP and surveillance systems. Routing protocols that discover and store more than one route in their routing table for each destination node are referred to as multipath routing protocols. In wireless scenarios, routes are broken due to node movement. Also, the wireless links used for data transmission are inherently unreliable and error prone. Therefore, multipath routing protocols are used to overcome the disadvantages of shortest path routing protocols. Multipath routing protocols are used to increase the reliability (by sending the same packet on each path) and fault tolerance (by ensuring the availability of backup routes at all times). It can also be used to provide load balancing, which reduces the congestion on a single path caused by bursty trafc. A survey
1550-445X/12 $26.00 2012 IEEE DOI 10.1109/AINA.2012.49

routes for each source-destination pair, a main route and a back-up route. The routes are discovered using two RREQ messages, each for one route. Whenever one route is broken, the other is used for data transmission and a RREQ is ooded to replace the broken route. This approach has two drawbacks: (i) MP-AODV has higher overhead than the traditional AODV because it requires one RREQ ooding for one path and additional RREPs for node-disjoint path and, (ii) the proposed approach is not able to nd all the available node-disjoint paths between a source and destination pair. Authors in [4] propose a scheme to nd all node-disjoint paths from source to destination. NDM-AODV also considers the residual energy of nodes while selecting the routes. The routing overheads to nd multiple paths are kept minimum by using Destination Source Routing (DSR) protocol like source routing in route discovery process. Periodic HELLO messages are used to maintain local connectivity for all active routes during the route maintenance phase. The main disadvantage of the proposed approach is that as the size of the network increases, the size of the RREQ and RREP messages also increases due to path accumulation function. Furthermore, the size of routing table at destination node also increases due to the storage required to store multiple paths. Several other implementations like AOMDV [5], AODVM [6] and AODVM-PES [7] also present multipath versions of AODV protocols. But unlike the above discussed methods, the multiple paths identied in these approaches are link-disjoint rather than node-disjoint. In all these methods, data transfer is started only after all multiple paths are discovered. This incurs initial delay in data transmission. Our proposed NDMP-AODV routing protocol addresses many problems associated with the previous node-disjoint multipath routing protocols. NDMP-AODV has back-up routes for active ows at all times which greatly reduces the delay. Approximately, above 80% control messages in AODV are RREQ messages. Our method minimizes the RREQ messages in network by nding all the available node-disjoint paths between a source and destination by ooding a single RREQ. NDMP-AODV is also combined with three different route maintenance processes which are described in the next section. III. P ROPOSED N ODE -D ISJOINT M ULTIPATH AODV ROUTING P ROTOCOL [NDMP-AODV] In this section, the proposed NDMP-AODV protocol is described. The main goal of NDMP-AODV is to nd all available node-disjoint routes between a source-destination pair with minimum routing overhead. To achieve this goal, NDMP-AODV protocol works in three phases: (i) Route Discovery Phase, (ii) Route Selection Phase and (iii) Route Maintenance Phase. A. Route Discovery Phase When a source node has a data packet to send, it checks its routing table for the next-hop towards the destination of the packet. If there is an active entry for the destination in the routing table, the data packet is forwarded to the next

hop. Otherwise, the route discovery phase begins. In route discovery phase, routes are determined using two types of control messages: (i) Route request messages (RREQs) and (ii) Route reply messages (RREPs). The source node oods the RREQ message into the network. Each intermediate node that receives a RREQ, checks whether it is a duplicate or a fresh one by searching an entry in the Seen Table. Seen Table stores two entries (i.e. source address and RREQ ooding ID ( )) that uniquely identies a RREQ message in the network. If an entry is present in the Seen Table for the received RREQ message, it is considered a duplicate RREQ message and discarded without further broadcasting. Otherwise, the node creates an entry in the Seen Table and updates its routing table for forward path before boardcasting the RREQ message.
Source IP Address Flooding ID Seen Flag

---

---

---

Fig. 1.

NDMP-AODV Seen Table structre

Type

Reserved

Prefix Size

Hop Count

Destination IP Address Destination Sequence Number Source IP Address Source Sequence Number

Broadcasting ID

Fig. 2.

NDMP-AODV RREP structre

In NDMP-AODV, only the destination node can send RREPs upon reception of a RREQ message. The intermediate nodes are forbidden to send RREPs even if they have an active route to destination. This is done so as to get the node-disjoint routes. In NDMP-AODV, the destination node has to send a RREP message for each RREQ received, even if the RREQ is a duplicate one. We change the data structure of Seen Table and RREP message as shown in Figures 1 and 2. In Seen Table, we add an extra eld that works as a ag known as seenflag. This ag is set to FALSE at start i.e. when an entry is rst inserted in the Seen Table after a node gets its rst RREQ message. The RREP messages initiated by destination node in NDMP-AODV contain one extra eld known as broadcast ID ( ). The route discovery method used to discover node-disjoint paths is shown in Figure 4. When a destination node receives a RREQ message, it creates the corresponding RREP message. The destination node copies the from the received RREQ message into the eld of sent RREP message. This

400

RREP is unicast towards the originator of the RREQ using the reverse path to construct the forward path. For every RREQ received (i.e. either rst or duplicate), the destination does the above mentioned process. When the intermediate nodes in the reverse path receive the RREP message, they check the seenflag value in their Seen Table. If the seenflag is set to FLASE, this indicates that this is the rst RREP message on the reverse path towards the source node. So, the intermediate nodes relay the RREP towards the source and reset the value of seenflag. When the intermediate node gets a RREP message for the same RREQ message it got earlier, the node simply discards the RREP message on the basis of seenflag value. Due to this, the intermediate nodes can only take part on any one route from the existing multiple routes. Algorithm 1 NDMR-AODV route discovery method when a node receives RREQ message = Node = Source Node = Destination Node = Intermediate Node = Source IP Address = Destination IP Address = broadcast id eld of RREQ = ooding id eld of RREP = FALSE //Initial value of seen ag seen table = FALSE = 0 if has data to send then if has route for then () end if else () () //insert entry in seen table to check for duplicates () end if if receives a RREQ message then if then = () //check for duplicate RREQs if then () //drop RREQ without reboardcasting else () //reboardcast RREQ end if else N is the destination = () //destination node send unicast RREP on forward route to create reverse route end if end if

Algorithm 2 NDMR-AODV route discovery method when a node receives RREP message if RREP then if = then = () //check and return the value of seenag from the seen table if then () // Insert rst route in routing table () //reset the value of seenag in seen table to detect duplicate RREPs () // forward RREP to next hop towards source else () //if multiple routes are stored at intermediate nodes () //drop the duplicate RREP to ensure the nding of node-disjoint routes end if else = () if = then () () else = () // count the numbers of active routes for destination in routing table if < n routes then () // insert secondary routes and sort them in acending hop count else () end if end if end if end if Algorithm 1 shows the procedure used by a node after getting a RREQ message. When a source node has a data packet to send, it checks its routing table for any active route available for destination. If an active route exists, data packet is forwarded to the next hop towards its destination. Else, it creates a RREQ message and inserts the entry in senttable. This is done to avoid re-sending RREQ messages before getting the RREP for the already sent RREQ. Each node also updates its Seen Table before broadcasting the RREQ message to avoid duplicate broadcasting. When a RREQ message is received by a node, the algorithm checks whether the node is a source, intermediate or destination node. If it is a source or intermediate node, RREQ message is processed in the same way as is done in the traditional AODV protocol. When a destination node receives the RREQ, it creates a RREP message and copies the value from RREQ into the extra eld provided in RREP. Destination node replies to every RREQ it receives to establish multiple routes. It does not check the received RREQ messages for duplicity as is done in AODV

401

protocol. When a node receives a RREP message during NDMRAODV route discovery process, Algorithm 2 is applied to discover multiple node-disjoint routes. The algorithm checks whether the node that receives a RREP message is an intermediate or source node. If it is an intermediate node, its seenflag status is checked from its Seen Table. A FALSE value of seenflag indicates that it is the rst RREP message that this node has received for this particular source-destination pair. If this is the case, the algorithm resets the value of seenflag corresponding to this source-destination pair and inserts this route as the primary route for the destination node. Then, the node forwards the RREP to the next hop towards source. On the other hand, if the value of seenflag is TRUE for this source-destination pair, we may or may not insert the route in the routing table as a secondary route, depending upon the route maintenance process used. This duplicate RREP message is then discarded to ensure that all the discovered routes are node-disjoint. If the node that receives the RREP is the source node, we insert the discovered node-disjoint path as primary or secondary, based on the value of seenflag and number of routes already present for this destination in routing table. Figure 3 shows the route discovery process of traditional AODV protocol. In Figure 4, we demonstrate with an example how the route discovery process in NDMP-AODV gets all node-disjoint routes between a source-destination pair. Suppose, node is the source node and node is the destination node. When node has data to send, it initiates the route discovery process by ooding RREQ in the network. Let us assume that destination receives its rst RREQ from intermediate node at time 1 and initiates the RREP1 message. RREP1 is unicast towards source by creating the reverse path DJMHES. When RREP1 is received by an intermediate node along the reverse route each intermediate node resets the value of seenflag in their Seen Table. Suppose, receives the rst duplicate RREQ message from at time 2 . Again node initiates a RREP2 for this duplicate RREQ and sends it back towards node through the same path it came to (i.e. SFCKPDD) to make the reverse route DAPKCFS. This helps to create a forward route towards node D. Finally, say at time 3 , node receives the third duplicate RREQ message from node . Node initiates RREP3 for this duplicate RREQ and sends it towards through . The RREP3 reaches node through . Node checks the value of seenflag for RREP3 before forwarding it to next hop. Node j determines that the seenflag is set to TRUE. So node considers RREP3 as a duplicate message and drops it. This helps to maintain the node-disjoint property of our method. B. Route Selection Process and Data Packet Transmission When source node has data packets to send and there is no route available in routing table, the node initiates the route discovery process. The source node starts data transmission as soon as it gets the rst route for destination node known

E H

F RREQ Messages RREP Messages Duplicate RREQ

Fig. 3.

Traditional AODV route discovery process

RREP2 RREP2 C F RREP2 RREP1 RREP1 S L B I Q O E H RREP1 K

RREP2 RREP2 P A RREP2 D

RREP1 M J

RREP1

RREP3 RREP3 R N

G RREQ messages Duplicate RREQ RREP messages

Fig. 4.

NDMP-AODV route discovery process

as primary route. All the other node-disjoint routes that are discovered will be stored in the routing table as secondary routes. After storing the primary route and an specied number of secondary routes in the routing table, all the other routes (if any) are not stored. All the other routes that are discovered after storing the primary and secondary paths can replace the existing secondary paths if they have lower hop count for destination as compared to existing ones. The route selection function works in such a way that whenever a route is required for data transmission, it always selects the primary route if it is available. If the primary route is not active, then the route selection function selects the route with lowest hop count from the available secondary routes.

402

C. Route Maintenance Process Route maintenance process is invoked when an active route is broken during completion of a data ow. We implement and analyze the performance of three route maintenance methods in case of route breaks. In the rst method, when the primary route is broken, transmission of data is continued using the secondary routes. To keep the secondary routes active while using the primary route, we increase the lifetime of each active secondary route after a xed amount of time. When all the secondary routes are also broken, the source starts a new route discovery process. In this way, we can minimize the routing overhead caused in nding and maintaining multiple routes. Because in this case, only one RREQ is used to nd all available node-disjoint paths as compared to one RREQ required for each path, as in the case of traditional AODV and other existing multipath extensions of AODV. This method when used with the proposed route discovery method, is referred to as NDMP-AODV in Section IV. In second route maintenance method, the source node starts the route discovery process as soon as it nds out that there is only one active path (i.e the one which is currently being used for data transmission) remaining in routing table. In this way, the source has routes for destination at all time. This greatly reduces the delay caused by the rerouting process which is triggered by a route break. But, this method increases the routing overhead. For example, if the source node stores two routes for destination, then every route break causes the ooding of a RREQ message to nd the backup route. If this is the case, then the routing overhead of this method are same or more than the AODV protocol. But if this method of route maintenance is used in low mobility networks where we can store more than two routes for destination node, it greatly reduces the routing overhead. This method when used with our proposed route discovery method, is referred to as BRAODV in Section IV. The third method is used in situations where the intermediate nodes on a route should also have to store multiple paths for destination. This can be done during our route discovery method without adding any extra routing overhead as shown in Figure 4. In Figure 4 we an see that node has two nodedisjoint routes for destination (i.e. JD and JND ). In this method, when an intermediate node detects route failure, it will select the secondary route (if available) from routing table and use that path for data transmission. This method decreases the RERR messages in network caused by intermediate nodes due to link breaks, thus increasing the network capacity. It also increases the PDR by not dropping the packets that are already on the broken route. This method, when used with our proposed route discovery process is referred as INMP-AODV in Section IV. IV. S IMULATION AND P ERFORMANCE E VALUATION In this section, we perform various simulations and analyze the results obtained. We incorporate changes in traditional AODV to make it node-disjoint multipath routing protocol. We evaluate and compare the performance of our proposed

route discovery method when used with three different route maintenance schemes and with traditional AODV. We use a commercial scalable network simulator called Qualnet to simulate our proposed methods.
TABLE I S IMULATION PARAMETERS

Parameters Values Simulator Qualnet 5.0 Simulation time 700sec Scenario Dimension 1200x1200 sq.meter Number of nodes 50 Transport protocol UDP Routing protocols AODV, NDMP-AODV, IM-AODV Mobility model Random way-point Pause Time 0 to 300 sec Path-loss model Two ray Radio type 802.11b Data rate 2mbps Video Sony Demo CIF(352x288) Video Codication H.264SVC Frame Size 22 to 27645 bytes Average Frame Size 1720 bytes Inter-packet Time 33 ms Number of Primary Routes 1 Number of Secondary Routes 1

A. Simulation Setup All the simulations are performed on a network that is modeled using the parameters listed in Table 1. The source nodes in the network are simulated for transmission of multimedia trafc in form of real time video stream. Constant bit rate (CBR) is used to produce background trafc in the network. The chosen video stream is the well-known CIF Sony Demo, whose size may be adequate for current PDAs and other mobile device displays. The ISO/IEC and ITU Joint Video Team (JVT) working group distributes reference software for SVC called JSVM (Joint Scalable Video Model), that is used for the implementation of the H.264/SVC [8] coding technique. The video trace le used to simulate the video sources during simulation is generated using JSVM 9.15. To evaluate the desired video streams, the video trace les obtained from [9] are converted into a format that is suitable for our simulator. At network layer, a single frame is transmitted using a single packet. In this way, we are able to stress the network with real-life video trafc instead of relying on VBR or CBR ows. Our evaluation is done over 700 sec because this amount of time is required to transmit the complete video stream from a source node. All sourcedestination pairs are selected randomly and all simulation

403

results presented in this paper are average results from 10 random simulation processes obtained by changing the seed value. B. Results and Analysis In this section, we discuss the results obtained from intensive simulations that have been performed to show the effectiveness of proposed route discovery and route maintenance methods. The simulation results include the average packet delivery ratio (PDR), average end-to-end delay (EED), percentage availability of backup routes and routing control overhead caused by route discovery and route maintenance processes. The effectiveness of proposed methods are checked against the effect of node mobility.
AODV NDMP-AODV 100 IM-AODV BR-AODV

Routing Control Overhead (%)

90 80 70 60 50 40 30 20 10 50 100 150 200 250 300

and plotted in Figure 5. The routing overhead is calculated by dividing the total number of routing control messages with the total number of packets in the network (i.e. control messages plus data packets). As we can see in Figure 5, AODV causes approximately 50% more routing overhead in moderate or low mobility networks (i.e. when node pause time is greater than 100 sec) as compared to NDMP-AODV. This is due to the fact that NDMP-AODV uses one RREQ ooding to calculate all node-disjoint routes as compared to AODV which uses one RREQ ooding for each route discovery. The number of RREP messages in NDMP-AODV is greater than AODV but they are very few in number because the RREPs are unicast towards source. Also, the intermediate nodes will not forward the duplicate RREPs. Low routing overhead saves the scarce network bandwidth, thus increasing the network capacity. The number of routes stored in routing table for a destination from the available node-disjoint routes greatly depends on the mobility of network. If the network mobility is high, the probability that the secondary route is expired with the primary route is high. As shown in Figure 5, BR-AODV has the highest routing overhead because only two routes for destination are stored in the routing table. Due to this, BRAODV has to ood the RREQ messages whenever any one route is broken to maintain the backup route at all times. In this case, the overhead for route maintenance is approximately more or equal to AODV protocol.
AODV NDMP-AODV IM-AODV BR-AODV

Pause Time (sec)

Fig. 5.

Routing control packet overhead with change in mobility

Average Packet Delivery Ratio (%)

100

95

90

AODV NDMP-AODV 0.2

IM-AODV BR-AODV

85

Average End-to-End Delay (sec)

80

0.15

75 50 100 150 200 250 300

Pause Time (sec)

0.1

Fig. 7.

Packet delivery ratio with change in node Mobility

0.05 50 100 150 200 250 300

Pause Time (sec)

Fig. 6.

Average End-to-End delay with change in node mobility

Figure 5 shows the overhead caused by routing control messages during route discovery process. Routing overhead created during transmission of one video stream are calculated

Effect of mobility on EED and PDR are shown in Figure 6 and Figure 7. The delay in NDMP-AODV is less as compared to other protocols. This is because NDMP-AODV keeps a backup routing path more than 50% of the time when the primary route fails with the lowest routing overhead. We can observe from Figure 6, that EED of all routing protocols decreases with increase in node pause time. NDMP-AODV EED again increases at the end of simulation due to increase in its PDR. Also, IM-AODV causes the highest delay because it uses the backup route from the point the link is broken which

404

may lead packets through longer routes. But it also increases the PDR of IM-AODV as shown in Figure 7. This is because the on-route data packets that are currently on the broken route are rerouted using the backup route from the point of route break. V. C ONCLUSION AND F UTURE W ORK In this paper, we propose a node-disjoint multipath routing method based on AODV protocol. The proposed route discovery method identies all the available node-disjoint routes using a single ooding of a RREQ message. This greatly reduces the routing overhead caused by route discovery and maintenance processes thus increasing the network capacity. Our method is suitable for low and moderate mobility networks as shown by the results in simulation section. Also, three different route maintenance methods are proposed and implemented to show the performance of our route discovery method. To reduce the initial delay, source node can send data as soon as it gets the primary route. Due to multiple routes stored in routing table backup routes are always available for continuous data transmission when the primary route is broken. Simulation results shows that NDMP-AODV is able to provide low end-to-end delay and high packet delivery ratio, while keeping low routing overhead. In future work, we will improve the route selection process of NDMP-AODV so that it can select routes that can satisfy user application requirements. Also, comparison with existing node-disjoint multipath routing protocols is left as a future work in this paper. We extend the proposed method to work efciently in high mobility networks by dynamically updating the backup route status. For transmission of multimedia trafc, a rate adaptation scheme should be combined with our multipath protocol. This is required for applications like video on demand, surveillance systems etc which have high transmission rates as compared to the available channel bandwidth. R EFERENCES
[1] Mohammed Tarique, Kemal E. Tepe, Sasan Adibi, and Shervin Erfani. Survey of Multipath Routing Protocols for Mobile ad hoc Networks. Journal of Network and Computer Applications, 32(6):1125 1143, 2009. [2] C. Perkins and E. Royer. Ad hoc on-demand distance vector routing. 1999. [3] Chang-Woo Ahn, Sang-Hwa Chung, Tae-Hun Kim, and Su-Young Kang. A node-disjoint multipath routing protocol based on aodv in mobile ad hoc networks. In Information Technology: New Generations (ITNG), 2010 Seventh International Conference on, pages 828 833, april 2010. [4] Shunli Ding and Liping Liu. A node-disjoint multipath routing protocol based on aodv. In Distributed Computing and Applications to Business Engineering and Science (DCABES), 2010 Ninth International Symposium on, pages 312 316, aug. 2010. [5] YuHua Yuan, HuiMin Chen, and Min Jia. An optimized ad-hoc ondemand multipath distance vector(aomdv) routing protocol. In Communications, 2005 Asia-Pacic Conference on, pages 569 573, oct. 2005. [6] Z. Ye, S.V. Krishnamurthy, and S.K. Tripathi. A framework for reliable routing in mobile ad hoc networks. In INFOCOM 2003. Twenty-Second Annual Joint Conference of the IEEE Computer and Communications. IEEE Societies, volume 1, pages 270 280 vol.1, march-3 april 2003. [7] Fubao Yang and Baolin Sun. Ad hoc on-demand distance vector multipath routing protocol with path selection entropy. In Consumer Electronics, Communications and Networks (CECNet), 2011 International Conference on, pages 4715 4718, april 2011.

[8] S. Rimac-Drlje, O. Nemcic, and M. Vranjes. Scalable video coding extension of the h.264/avc standard. In ELMAR, 2008. 50th International Symposium, volume 1, pages 9 12, sept. 2008. [9] Video Trace:. http://trace.eas.asu.edu/.

405