This action might not be possible to undo. Are you sure you want to continue?
5, August 2010
Convergence Time Evaluation of AODV and AODV+G in MANETs
Annapurna P Patil
Department of Computer Science and Engineering, M.S. Ramaiah Institute of Technology,Bangalore-54,India. firstname.lastname@example.org email@example.com
Abstract -Wireless mobile ad-hoc networks are characterized as networks without any physical connections. In these networks there is no fixed topology due to the mobility of nodes, interference, mulitpath propagation and path loss. Hence a dynamic routing protocol is needed for these networks to function properly. Many routing protocols have been developed for accomplishing this task. Selecting most appropriate routing protocol for a particular network scenario is the critical issue. Most attempts made at evaluating these algorithms so far have focused on parameters such as throughput, packet delivery ratio, overhead etc. An analysis of the convergence times of these algorithms is still an open issue. The work carried out fills this gap by evaluating the algorithms on the basis of convergence time. In this paper we present and examine the convergence time evaluation of routing protocols AODV and AODV+G . The algorithm performances are compared by simulating them in ns2. Tcl is used to conduct the simulations, while perl is used to extract data from the simulation output and calculate convergence time. After extensive testing we observed that AODV+G converged well in all situations than AODV. The paper also evaluates the algorithms using the rudimentary metrics-throughput and packet delivery ratio. Keywords- Routing Protocols, MANETS, Convergence Time.
I. INTRODUCTION A Mobile Ad-Hoc Network (MANET) is a selfconfiguring network of mobile nodes connected by wireless links, to form an arbitrary topology. The nodes are free to move randomly. Thus the network's wireless topology may be unpredictable and may change rapidly. Minimal configuration, quick deployment and absence of a central governing authority make ad hoc networks suitable for emergency situations like natural disasters, military conflicts, emergency medical situations etc . Every device in a MANET is also a router because it is required to forward traffic unrelated to its own use. Almost every year, the world is struck by numerous catastrophic natural disasters, such as earthquake, hurricane, typhoon, tsunami, etc. In such a situation communication systems, fixed or mobile, were usually down due to various reasons. The loss of communication systems as well as information networks made the rescue operation extremely difficult.
WiFi-ready notebook PCs( MANET nodes) owned by rescue volunteers themselves to construct a MANET to support such a need. MANET can be classified based on the communication pattern or the devices used, the variants of MANETs on the type of devices are sensor and ad hoc networks. Routing is one of the critical issue in MANET. Selecting the energy efficient routing protocols improves the performance of the communication. The routing protocols are classified into three types. Proactive protocols maintain routing information for all the destinations, and keep updating this information through periodic updates, an example for this protocol is DSDV,OLSR. Reactive protocols don’t maintain information for all the destination, rather they discover the route to a destination on demand, an example for this protocol is AODV. Hybrid protocols attempt to combine the advantage of both proactive and reactive protocols, an example for this protocol is TORA, ZRP, MPOLSR. AODV+G reduces unnecessary traffic will effectively improve the efficiency of those mobile nodes in network. AODV and AODV+G protocols performs differently under different network scenarios. One protocol might perform better than others in specific situation. These protocols are compared in terms of convergence time to uncover in which situations these types of algorithms have their strengths and weaknesses. II. RELATED WORK There are many other works which are related to our work in evaluating routing algorithms.  AODV and AODV+G has been compared in terms of Average delay, Packet delivery ratio, Normalized routing load and Routing load reduction, but not in terms of convergence time.  AODV and DSDV has been compared in terms of convergence time. Many papers have compared AODV with other routing algorithms. In  AODV and DSDV have been compared with average throughput, packet loss ratio, and routing overhead as the evaluation metrics,  has compared AODV and DSDV in terms of delay and drop rate,  compares AODV and DSDV in terms of throughput, packets received, delay and overload. Similarly,  compares
175 http://sites.google.com/site/ijcsis/ ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 5, August 2010
AODV, DSDV and DSR in terms of throughput, delay, drop rate. III. PROTOCOL SPECIFICATION This section gives the small presentation of two protocols we evaluate in this paper. A. AODV The Ad-hoc On-Demand Distance Vector (AODV) routing protocol is designed for use in ad-hoc mobile networks. AODV is a reactive protocol: the routes are created only when they are needed. It uses traditional routing tables, one entry per destination, and sequence numbers to determine whether routing information is up-to-date and to prevent routing loops. An important feature of AODV is the maintenance of time-based states in each node: a routingentry not recently used is expired. In case of a route is broken the neighbours can be notified. Route discovery is based on query and reply cycles, and route information is stored in all intermediate nodes along the route in the form of route table entries. The following control packets are used: routing request message (RREQ) is broadcasted by a node requiring a route to another node, routing reply message (RREP) is unicasted back to the source of RREQ, and route error message (RERR) is sent to notify other nodes of the loss of the link. HELLO messages are used for detecting and monitoring links to neighbours. B. Gossiping & AODV+G The basic gossiping protocol is simple. A source sends the routing request with probability 1. When a node first receives a routing request, with probability p it broadcasts the request to its neighbors and with probability 1 – p it discards the request; if the node receives the same rout request again, it is discarded. Thus, a node broadcasts a given route request at most once.  proposes GOSSIP(p,k,m), an extension to the basic gossiping, and suggests that: A node broadcasts with probability 1 for the first k hops before continuing to gossip with probability p. If a node with n neighbors receives a message and does not broadcast it, but then does not receive the message from at least m neighbors within a reasonable timeout period, it broadcasts the message to all its neighbors . Hass et al. implements GOSSIP(p,k,m) in Ad Hoc On-Demand Distance Vector protocol (AODV) , a typical and well-studied on-demond routing algorithm suited for mobile nodes routing in ad hoc network. We refer this gossipbased AODV as AODV+G. The experiments in  shows that gossiping can reduce control traffic up to 35% when compared to flooding and the most significant performance of GOSSIP is achieved by taking p=0.65, k=1 and m=1. In AODV+G, if the expanding-ring search with a smaller radius fails, rather than flooding to the whole
network, here GOSSIP3(.65,1,1) is used. The timeout period of GOSSIP3 should be big enough to allow neighboring nodes to gossip. The NODE_TRAVERSAL_TIME parameter of AODV is a conservative estimate of the average one hop traversal time for packets that includes queuing delays, interrupt processing times and transfer times. GOSSIP3 is not used in the expanding-ring search with a smaller radius, since flooding is more efficient than gossiping for zone with small radius because of the back-propagation effects. The variant of AODV that uses GOSSIP3 is called AODV+G.
IV SIMULATION AND PERFORMANCE ANALYSIS A. Environment and Assumption Simulator chosen : The proposed algorithms are simulated on NS2(version 2.33). NS2 is popularly used in the simulation of routing and multicast protocols, among others, and is heavily used in ad-hoc networking research. ns supports an array of popular network protocols, offering simulation results for wired and wireless networks alike. It can be also used as limited-functionality network emulator. It was necessary to use available implementations of algorithms rather than implement them freshly ourselves, as it is important for the acceptance of an evaluation that the implementation used for evaluation has been scrutinized and accepted as correct by the community. Else the evaluation results will not be accepted as doubt will exist about the correctness of the implementation of the algorithms Algorithms chosen : Here in this paper we have selected to simulate and evaluate the performance of AODV and AODV+G protocols. AODV is a reactive routing protocol and AODV+G is variant of AODV routing protocol with GOSSIP3. Further experiments can be built based on the results of this project, to compare convergence time performance of algorithms within the same category as well. Mobility model : The Random Waypoint model is the most commonly used mobility model in research community. At every instant, a node randomly chooses a destination and moves towards it with a velocity chosen randomly from a uniform distribution [0,V_max], where V_max is the maximum allowable velocity for every mobile node. After reaching the destination, the node stops for a duration defined by the 'pause time' parameter. After this duration, it again chooses a random destination and repeats the whole process until the simulation ends. To create Mobile node Movement Scenario files, the command line that needs to be run under directory : nsallinone-2.33/ns-2.33/indep-utils/cmu-scen-gen/setdest : ./setdest [-n num_of_nodes] [-p pausetime] [-s maxspeed] [-t simtime] [-x maxx] [-y maxy] > [output-file].
176 http://sites.google.com/site/ijcsis/ ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 5, August 2010
Traffic pattern : Moreover, traffic sources may generate packets at constant bit rate (CBR), or at variable bit rate (VBR). The CBR class is commonly used for voice and data services. In this context, the data rate and the delay remain constant during the packet transmission. More particularly, CBR traffic sources provide a constant flow of data packets of 512 bytes with a transmission rate of 4 packets per second. All CBR traffic scenarios are generated using cbrgen.tcl in NS-2 To create CBR traffic scenario files, under directory :ns-allinone-2.33/ns-2.33/indep-utils/cmu-scen-gen/cbrgen.tcl ./ns cbrgen.tcl [-type cbr|tcp] [-nn nodes] [-seed seed] [-mc connections] [-rate packet/second for one connection] > [output-file]. Network scenario : The simulations are conducted using the network simulator ns2 . Random Waypoint mobility model is used. The physical layer simulates the behavior of IEEE 802.11 (as included with ns2). Each node has a radio range of 250 meter, and uses TwoRayGround as the radio propagation model. All the scenarios are based on the following basic parameters: cbr (constant bit rate) traffic topology of size 500 m x 500 m maximum speed of each node 20 m/s simulation time 180s transmission rate (packet rate) 10 m/s The number of nodes is varied in the range [10,100] in steps of 10 (to represent 10 node densities). Pause time is varied in the range [0,180] in steps of 20 (to represent 10 pause times). B. Performance Metric A trace file contains a lot of information which may not be required to analyze the performance of the protocol. We are always interested in some amount of information that is sufficient to predict the efficiency of the protocol. The following performance metric is needed to be taken into consideration in order to analyze and compare the performance of AODV and AODV+G Convergence Time : In , convergence time has been defined as the time between detection of an interface being down, and the time when the new routing information is available.  defines a route convergence period as the period that starts when a previously stable route to some destination becomes invalid and ends when the network has obtained a new stable route for. Similarly, we define convergence time as the time between a fault detection, and restoration of new, valid, path information.  calculates convergence time in the IP backbone. The authors arrive at the value of convergence time by deploying entities called ‘listeners’, which listen to every link state PDU being sent by the is-is protocol. The time when the first ‘adjacency down’ packet is observed is taken as the time
of detection of an interface being down. This failure event is said to end when the listener receives link state PDUs from both ends of the link. We arrive at the convergence time by measuring the interval between the detection of route failure and successful arrival of a packet at the destination over the newly computed route. This includes not only the routing convergence time, but also the time taken for the packet to traverse the network from the source to the destination over the newly discovered path. Since this is a comparative analysis, and both the routing protocols use shortest distance with number of hops as the metric for distance calculation, both protocols will arrive at the same new route, and the time taken to reach the destination over this new route will be the same (since all physical characteristics are the same). Hence this extra time measured does not affect the comparative analysis. In any case, the time taken for a packet to travel from the source to the destination is negligible when compared to the time taken for the algorithm to discover the new route, either through route request – route reply sequences as in reactive protocols, or by waiting for an update that contains new route information as in proactive protocols. Also, this automatically verifies that the new path calculated is correct. The cycle of invalidation of old path and discovery of a new path might occur many times, and for many sourcedestination pairs over the course of the simulation. Hence the average value of these times is taken as the convergence time of that algorithm for that scenario. This procedure has been carried out in perl. Throughput : If y number of packets delivered within t time at a node then the throughput at the node could be defined as y/t. By definition, the throughput needs to be calculated at the bottleneck node, not sender. For the throughput calculation, in general divide the successfully received packets by the simulation time will give the answer. In the trace file there are different levels of received packets such as the RTR or AGT level. The packets received by the node in its AGT level will be the real received packets. Here these packets are filtered from the trace file using perl script. Packet Delivery Ratio : The ratio between the number of packets successfully received by the application layer of a destination node and the number of packets originated at the application layer of each node for that destination.
V. EXPERIMENTAL RESULTS Graphs are one of the ways to analyze and compare the results of the trace file. Other methods can also be used for comparison like tabular form showing required output data of the trace file. Simple MS Excel or MATLAB also work for plotting graphs. In this paper the graphs are plotted using xgraph in NS2.
177 http://sites.google.com/site/ijcsis/ ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 5, August 2010
In order to be able to cover most if not all the types of scenarios the algorithms might face, we varied both the node density (number of nodes) and the node mobility (pause time). The node density (number of nodes) was varied in the range [10,100] in steps of 10 (10 different node densities). The upper limit of this range was chosen to be 180 because the simulation time is 180s in all the cases. Thus a pause time of 180 implies that the nodes pause in their initial positions for 180 seconds – the entire duration of the simulation. Hence this represents the case where nodes are completely static. Similarly, pause time 0 represents very high mobility where the nodes are in constant motion. Thus we tested each algorithm over 10 node densities x 10 pause times = 100 scenarios. A. Convergence Time Convergence time of AODV and AODV+G is calculated using perl script. This script parses trace file created by simulating AODV and AODV+G algorithms to calculate convergence time. In each graph, the node density is fixed and the pause time is varied. Figure 3: 30 nodes, varying pause time
Figure 4: 40 nodes, varying pause time Figure 1: 10 nodes, varying paused time
Figure 2: 20 nodes, varying pause time
Figure 5: 50 nodes, varying paused time
178 http://sites.google.com/site/ijcsis/ ISSN 1947-5500
Figure 6: 60 nodes, varying paused time
Figure 9: 90 nodes, varying paused time
Figure 7: 70 nodes, varying paused time Figure 10: 100 nodes, varying paused time Based on the above figures it is found AODV+G convergence time is less than AODV in all assumed network scenarios. It is also found that as the pause time increases the convergence time of AODV+G decreases. Convergence time of both AODV and AODV+G increases as the node density increases. B.. Throughput Here node density is varied from 10 to 100 in steps of 10 nodes. For each node density both the algorithms are simulated with varied paused time from 0s ts 180s in steps of 20s. Average throughput in each node density is taken and the graph is plotted as show in figure 12.
Figure 8: 80 nodes, varying paused time
179 http://sites.google.com/site/ijcsis/ ISSN 1947-5500
VI. CONCLUSION AODV and AODV+G mobile Ad-hoc routing protocols have been presented and evaluated using well know network simulator NS2( version 2.33 ). AODV+G is gossip based AODV, here GOSSIP3(.65,1,1) is used. These two protocols are evaluated using the network performance metric convergence time. Here we observed that AODV+G converged well than compared to AODV in all assumed network scenarios. We also noticed that with the very low node density throughput and packet delivery ratio of AODV is more than AODV+G. With node density more than 30 nodes AODV+G performs better than AODV. We can extend our work to compare the performance of Adaptive Gossipbased Ad Hoc Routing (AGAR) with Gossip-based Ad Hoc Routing (AODV+G) using convergence time, throughput and packet delivery ratio. ACKNOWLEDGMENT We wish to acknowledge our Principal Dr K Rajanikanth, M. S .Ramaiah Institute of Technology,Bangalore-54 and Professor and Head of the Department at CSE Prof .V. Muralidharan for their encouragement which helped us produce this work.. VII. REFERENCES
 Charles E.Perkins a nd Pravin Bhagwat, “Highly Dynamic Destination-Sequenced Distance-Vector Routing (DSDV) for Mobille Computers ” In Proceeding of SIGCOM ’ 94 Conference on Communications Architecture, protocols and Applications August 1994. “Ad hoc On Demand Distance vector (AODV) Routing protocol”, RFC 3561, july 2003. Jing Xie, Luis Girons Quesada and Yuming Jiang “A Threshold-based Hybrid Routing Protocol for MANET” Department of Telematics, Norwegian University of Science and Technology. 2008 Nicklas Beijar, “Zone Routing Protocol (ZRP)”, Networking Laboratory, Helsinki University of Technology. 1998 [ns] MANET : TORA Routing Protocol at http://firstname.lastname@example.org/msg05173.html. Jiazi YI, Sylvain David, Asmaa Adnane, Benoit Parrein, “ MultipathLSR. Simulation and Testpath”, 5th OLSR Interop/Workshop, Vienna, Austria (2009) Z. J. Haas, J. Halpern, and L. Li. Gossip-based ad hoc routing. IEEE Proceedings of INFOCOM, 2002. Annapurna P Patil, Narmada Sambaturu, Krittaya Chunhaviriyakul, “Convergence Time Evaluation of Algorithms in MANETs”, (IJCSIS) International Journal of Computer Science and Information security, julyb2009. R. Khalaf, A. El-Haj-Mahmoud, and A. Kayssi (Lebanon) “Performance Comparison of the AODV and DSDV Routing Protocols in Mobile Ad Hoc Networks”, From Proceeding (371) Communication Systems and Networks – 2002.
Figure 12: Throughput AODV Vs AODV+G
From the above figure we observed that AODV with the low density performance well than AODV+G. As the node density increases AODV+G throughput increases. With the node density more than 30 nodes the throughput of AODV+G is almost double than AODV. C. Packet Delivery Ratio Here node density is varied from 10 to 100 in steps of 10 nodes. For each node density both the algorithms are simulated with varied paused time from 0s ts 180s in steps of 20s. Average of packet delivery ratio in each node density is taken and the graph is plotted as show in figure 11.
  
Figure 11: Packet Delivery Ratio AODV Vs AODV+G From the above figure we observed that AODV with the low density performs well than AODV+G. With the node density more than 30 nodes packet delivery ratio of AODV+G is more than AODV.
180 http://sites.google.com/site/ijcsis/ ISSN 1947-5500
Mahdipour , Ehsan Aminian, Mohammad Torabi,  Ebrahim Mehdi Zare, "CBR Performance Evaluation over AODV and DSDV in RW Mobility Model, " Computer and Automation Engineering, International Conference on, pp. 238-242, International Conference on Computer and Automation Engineering (iccae 2009), 2009.  Muazzam Ali Khan Khattak, Khalid Iqbal, Prof Dr. Sikandar Hayat Khiyal, “ Challenging Ad-Hoc Networks under Reliable & Unreliable Transport with Variable Node Density”, Journal of Theoretical and Applied Information Technology, 2005-2008.  REN Wei, YEUNG D.Y, JIN Hai1” TCP performance evaluation over AODV and DSDV in RW and SN mobility models” (School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan 430074, China) (Department of Computer Science, Hong Kong University of Science and Technology, Hong Kong, China) 2005.  The Network Simulator - ns-2 at http://www.isi.edu/nsnam/ns/  Jaspreet kaur and Cheng Li, “ Simulation and analysis of Multicast protocols in Mobile Ad Hoc Networks using NS-2”, Memorial University of Newfoundland St.John’s, Newfoundland, A1B 3X5, Canada, 2007  G. Iannaccone, C.-N. Chuah, R. Mortier, S. Bhattacharyya, and C. Diot, “Analysis of link failures over an IP backbone,” in ACM SIGCOMM Internet Measurement Workshop (IMW), Nov. 2002.  D. Pei, X. Zhao, L. Wang, D. Massey, A. Mankin, S. F. Wu, and L. Zhang, “Improving bgp convergence through consistency assertions, ” in Proc. of IEEE INFOCOM, 2002.  Zygmunt J. Haas, Senior Member, IEEE, Joseph Y. Halpern, Senior Member, IEEE, and Li (Erran) Li, Member, IEEE, “Gossip-Based Ad Hoc Routing”, IEEE/ACM TRANSACTIONS ON NETWORKING, VOL. 14, NO. 3, JUNE 2006
The authors are Faculty and Post graduate Student at M S Ramaiah Institute of Technology, Bangalore working in the area of performance evaluation of routing algorithms at the R&D labs, Department of Computer Science and Engineering.
181 http://sites.google.com/site/ijcsis/ ISSN 1947-5500