distributed multicast routing for mobile ad hoc networks

7/27/2019 Distributed Multicast Routing for Mobile Ad Hoc Networks

1/6

Proceedings of National Conference on Emerging trends in VLSI, Embedded and Nano Technologies,

January 27 - 28, 2011.

461

Distributed Multicast Routing For Mobile Ad Hoc Networks

Ms.S.Priya/PG Student

Dept of Computer Science and Engineering,

Mepco Schlenk Engineering College,

Sivakasi, Tamil Nadu

E-mail: [email protected]

Mr.C.Balasubramanian/Assistant Professor

Dept of Computer Science and Engineering,

Mepco Schlenk Engineering College,

Sivakasi, Tamil Nadu

E-mail: [email protected]

AbstractRouting over mobile ad hoc networks is

complicated by the considerations of energy efficiency.

Most of the mobile devices operate on battery and hence

power consumption becomes an important issue. This

paper targets power-aware routing when the network

topologies and data traffic may change quickly in anunpredictable way. In this paper, a distributed

algorithm called Maximum Residual Multicast

Algorithm (MRMA) is proposed to maximize the

minimum residual energy of all the nodes for each

multicast without collecting the global information. A

transient multicast tree is established on demand based

on the autonomous decisions of intermediate nodes.

Based on the proposed algorithm, a pure source initiated

on-demand routing protocol is developed with special

designs to adapt to large-scale mobile ad hoc networks

with energy-efficiency considerations.

KeywordsMaximum residual rou ting, routi ng protocols,

mobil e ad hoc networks.

I. INTRODUCTIONA mobile ad hoc network (MANET), sometimes called a

mobile mesh network, is a self-configuring network ofmobile devices connected by wireless links. Routingbecomes highly challenging because of the dynamic natureof network topologies and critical energy consideration. Theproblem is more complicated by the replacement of multipleunicasts with a multicast. Example applications are inadvertisement, video/audio conferencing, shopping malls,distance learning, tourist information distribution and taxidispatching .

In the recent years, many excellent routing protocolshave been proposed for mobile ad hoc networks, e.g.,[11],[12],[13]. Each of them tried to optimize some routingand performance metrics for different application scenario.In order to save network bandwidth, multicast protocolswere also widely explored, e.g.,[5],[6],[7]. Among theseMAODV [4], ODMRP [8], and DDM [9] provide anexcellent solution. MAODV discovers tree-based routes ondemand using a broadcast route discovery mechanism.ODMRP, which is mesh based protocol, uses a forwardinggroup concept (only a subset of nodes forwards the

multicast packets). A soft state approach is taken inODMRP to maintain multicast group members. No explicitcontrol message is required to leave the group. In DDM,each source is responsible for the maintenance of multicastgroup.

Routing over MANETS is complicated by theconsideration of energy efficiency, while the shortest path isnot favored. Many results are presented in the literature,e.g., [1], [2]. In recent years, power-aware routing hasreceived lots of attention and yielded a class of fundamentaloptimization problems over various routing metrics. One ofthe well known metric of power-aware routing is Minimum-Energy routing, which tries to minimize the energyconsumed by each packet. Another popular metric is basedon Maximum Lifetime routing, which maximizes the firstnode failure time and also the partition time [2], [3]. Most ofthe existing works rely on the knowledge of the certainglobal information such as remaining energy and also theminimum transmission between nodes. The maintenance ofthe global information is highly challenging in designing theprotocol, because of the difficulty and cost in themaintenance of up-to-date information.

In this paper, we are interested in Maximum-ResidualRouting, where the minimum residual energy of each nodeis maximized for each multicast and the network lifetimeincreases. There are algorithms which relay on the globalinformation of each node. Unlike the past work, we firstpropose a distributed algorithm called Maximum-ResidualMulticast, which derives a loop-free route. Based on theproposed algorithm, we then develop a source initiated ondemand routing protocol which is referred to as Maximum-Residual Multicast Protocol (MRMP), is adaptable todynamic changes in network topologies.

The rest of this paper is organized as follows: Section IIprovides the formulation of problem. In Section III, adistributed algorithm is proposed and then the routingprotocol is designed based on the algorithm. Simulationresults and analysis are discussed in Section IV and SectionV provides our conclusion.

II. PROBLEM FORMULATIONEach multicast request might need to be partitioned into

multiple sessions. The problem is formulated as the


2/6


January 27 - 28, 2011.

462

maximization of the minimum remaining energy of nodes inthe network after each multicast session, where theremaining energy of a node after each multicast is referredto as its residual energy. The goal is to derive a route so asto maximize the minimum residual energy of nodes in thenetwork without collecting and storing the detailed topologyand the remaining energy information of the whole networkat any node.

The network model under considerations can beformulated as a directed graph G=(V, E), where each nodeu V is associated with its remaining amount of energy,denoted by (u). Different nodes consume different energyin packet transmissions with different communicationranges. Each directed edge (u,v) E is associated with aweight w(u,v) to denote the amount of energy needed for anode u to transmit one session of data packets to anothernode v, and a constant (v) denotes the energy consumptionof receiving one session of data packets for node v.

Definition : The Maximum-Residual Multicast Problem

Suppose that there is an ad hoc network G = (V,E),where each node u V is associated with an amount (u)of its remaining energy and a constant amount (u) ofenergy in receiving one session of data packets, and eachdirected edge (u,v) E is associated with a weight w(u,v) totransmit one session of data packets from u to v. Given asource s V and a destination set R V the problem is to

find a multicast tree T G that is rooted at s and includes

all nodes in R such that (T) is maximized. The tree isreferred to as a maximum residual multicast tree.

III. A MAXIMUM-RESIDUAL MULTICAST PROTOCOLWe first propose a distributed algorithm in Section A, to

derive a multicast tree with the best energy efficiency,where each node makes its own decision autonomously. Arouting protocol is then developed in Section B as arealization based on the proposed algorithm.

A. A Distributed AlgorithmA distributed methodology and its implementation are

proposed to resolve the maximum-residual multicastproblem. Based on each multicast tree T derived by theproposed algorithm, every node is able to adjust its powerlevel in packet transmissions so that the residual energy overa network G = (V,E) is maximized for a given multicastsession S. Let the source and the destination set of S bedenoted by s and R, respectively. Given each node v V

under considerations, [v] and m[v] are used to keep trackof its predecessor and estimation on the residual energy overa path from s to itself during the execution of the algorithm,respectively.The proposed algorithm is referred to as theMaximum- Residual Multicast Algorithm (MRMA).

For source s

If s has a session S of data packets to multicast to nodesin R then

Create an entry indexed by (s,S) at s;

m[s] (s);[s] NIL;Broadcast msg (s, S, (s), m[s], 0) to its neighbors.

When a session S is initiated, MRMA is invoked. Anentry associated with S is created at s by setting m[s] and[s] as (s) and NIL respectively. Because s is the source,the residual energy over the path of only s remains as (s)(i.e., its remaining amount of energy). A control message isthen broadcasted to all of the neighbors of s. For each nodeu, its control message carries the source identification s, thesession number S, its remaining energy (u), an estimationm[u] on the residual energy over a path from s to itself, andthe energy consumption (u) for u in receiving one session.Here s does not need to consume energy for the reception,and thus, the last field in its control messages is set as 0.

For a node v other than s

if v receives msg (s, S, (u), m[u], (u)) from a neighbor uthen

if no entry is indexed by (s,S) at v thenCreate an entry indexed by (s,S) at v;m[v] 0;[v] NIL;

if m[v] < min {m[u], (u) w(u,v)(u), (v) (v) }then

m[v] min{ m[u], (u) w(u,v)(u), (v) (v) }[v] u

Broadcast msg (s, S, (v), m[v], (v)) to all of itsneighbors;

When a node v receives a control message from aneighbor u , it should first check up whether an entry existsfor the corresponding session. If not, an entry is created bysetting the initial values of as 0 and NIL, respectively. Thiswill result in the update of m[v] and [v] and the broadcastof a control message to all of its neighbors if a path from sto v (through u) with higher residual energy is found i.e., uis a better predecessor of v in terms of residual energymaximization if the current value of m[v] is smaller thanmin{ m[u], (u) w(u,v)(u), (v) (v) },where is theestimate on the residual energy over the current path from sto u in T, ((u) w(u,v)(u)) denotes the residual energyof u over a path through the edge (u ,v), and ((v) (v) )denotes the residual energy of v. Node v has all of theinformation needed in making the decision, except theamount of energy consumed by u to transmit the packets ofthis session to v, i.e., w(u,v). This information cannot becarried in the control message because u has no knowledgeabout this information.

B. Protocol DesignThe MRMP, a realization of routing algorithm MRMA,

is a pure source-initiated on-demand routing protocol, whichestablishes routes if and only if they are desired by sources.MRMP uses a broadcast route discovery mechanism, asused in other on-demand routing protocols [4], [8], withspecial designs to adapt to large-scale mobile ad hoc


3/6


January 27 - 28, 2011.

463

networks with energy-efficiency considerations. In MRMP,a route is established by autonomous decisions ofintermediate nodes, instead of being determined by thesource with global information. Furthermore, no noderegularly maintains routes to others. No periodic controlmessage exists for neighborhood and group management,and no control message is initialized for route repairs. Aroute (once established) is considered valid within sometime interval (relative to the moving velocity) because it isdetermined on demand based on the network status at thetime being, rather than on any information collected earlier.

Since MRMA is realized by MRMP, control messagesand table entries of MRMP directly correspond to theircounterparts of MRMA. There are three major stages inMRMP:

1. Route Discovery2. Route Establishment.3. Data Forwarding.

Whenever a source needs a route, the source requests fora route discovery within the network so that each node

decides its predecessor. The destination nodes and theirancestors then inform their predecessors of the proper powerlevels during route establishment. Then data packets areforwarded by nodes at proper power levels on theestablished route.

In MRMP, two tables are maintained at each node: TheGroup Table and the Route Table. The Group Table is usedto maintain the group information of the node, where eachsource represents a group. A node becomes a destination ofa group if the corresponding source is added into its GroupTable. Note that which groups a node joins is maintainedby each node itself autonomously. No control message isneeded to inform others of changes in its Group Table. TheRoute Table is used to record the proper power level to

transmit the data packets of each session. This table consistsof routing entries, where each entry is associated with asession (i.e., the counterpart of an entry in MRMA). Anentry remains valid until it is removed because of the tablespace limitation. The fields of an entry are as follows:

Source ID (i.e., s); Session Number (i.e., S); Transmission Power; Remaining Energy (i.e., ); Estimate (i.e., m); Predecessor (i.e., ); Adjust Ratio; Membership Flag Entry Status.

An entry is indexed by a unique pair of Source ID andSession Number, where an example implementation ofSource ID is the IP address of the source that originates thissession, and Session Number can be a sequence numbermaintained in the source to distinguish different sessions.Transmission Power records the proper power level of thenode to transmit the data packets of the associated session.This transmission power is a replacement of the role of NextHop in a traditional route table, and it is set during the routeestablishment. Remaining Energy is set as the amount of the

remaining energy of the node when this entry is created (andit remains the same thereafter). Estimate, Predecessor, andAdjust Ratio, are used to keep track of a temporary decisionin route discovery and are subject to changes. Estimate andPredecessor are set according to MRMA. Adjust Ratioindicates the ratio of the maximum transmission power ofits current predecessor such that it itself can receive datapackets from the predecessor successfully.

Membership Flag indicates the relationship between thenode and the session associated with this entry. This flagcan be set as either IN_GROUP (i.e., a destination of thesession), ON_TREE (i.e., an intermediate node that helps toforward the session), or NO_RELATION (i.e., a nodeunrelated to the session). Entry Status denotes the currentstage as RTF_DISCOVERY, RTF_ESTABLISHMENT orRTF_READY. Furthermore, each entry is associated withtwo timers, referred to as dtimer and etimer. They are usedto trigger stage changes of the entry.

1. Route Discovery StageA route discovery procedure is invoked when a source

has data packets to send. The source creates an entry in itsRoute Table with Membership Flag and Entry Statusinitialized as ON_TREE and RTF_DISCOVERY,respectively. The dtimer associated with this entry is thenactivated. Route discovery begins with a broadcast of arequest message (REQ) from the source to all of itsneighbors with its maximum transmission power. The dataframe of an REQ contains the following fields:

< source ID; session number;

packet size; number of packets;

remaining energy; estimate >The pair of the first two fields is used to identify the

REQs employed for a specific session. Packet size andnumber of packets, respectively, denote the number of bitsin a data packet and the number of packets in the session.The last two fields are used to carry the values of theremaining energy and the estimate recorded in theassociated entry.

Figure 1. Route Discovery

When a node v receives an REQ from a neighboringnode u, a corresponding entry is created in the Route Table


4/6


January 27 - 28, 2011.

464

of v, if it does not exist. The routing information is thenextracted from the REQ to test whether Estimate can beincreased. If it can be increased, Estimate, Predecessor, andAdjust Ratio are updated, and then the REQ isrebroadcasted to neighbors with the last two fields replacedby the Remaining Energy and the new Estimate of node v.This process is repeated until the dtimer expires.

2. Route Establishment Stage

The destination nodes and their ancestors then informtheir predecessors of the proper power levels during routeestablishment. As the dtimer of an entry expires, the etimeris activated and the Entry Status is set asRTF_ESTABLISHMENT. For those REQs (devoted to thissession) that arrive late are simply discarded. Routeestablishment is to let each node (if needed) inform itspredecessor of the Adjust Ratio kept in its Route Table sothat the predecessor will use a proper power level totransmit the session. Each destination or intermediate nodesends exactly one reply message (RPY) to inform itspredecessor. After a destination (/intermediate) node sends

its RPY, it changes its Membership Flag to IN_GROUP(ON_TREE). The data frame of an RPY contains thefollowing fields:

< source ID; session number; adjust ratio >No handshaking mechanism is provided for multicasting

or broadcasting in the MAC layer. In order to prevent theonly one RPY sent by a node from collision, the destinationfield in the IP header of the RPY is set as the address of thepredecessor, rather than the broadcast address used inREQs.

In this way, each RPY is considered as a one-hop unicastpacket by the MAC layer, and a handshaking mechanism,such as RTS/CTS, can be used to reserve the medium forthe duration required to send the RPY, thus, mitigatingcollision. Even if collision occurs, the MAC layer willautomatically retransmit the RPY to further improve theprobability of successful reception.

Figure 2. Route Establishment

During route establishment, a node may receive multipleRPYs and the maximum value of Adjust Ratio must be kept.When the etimer expires, the field Transmission Power inthe corresponding entry is set as its maximum transmissionpower level multiplied by the kept Adjust Ratio.

3. Data Forwarding Stage

When the etimer of an entry expires, the Entry Status isset as RTF_READY and the routing entry is consideredactive. When an entry of a source is active, the sourcebegins to transmit the data packets of the associated sessionusing the Transmission Power in the entry.

Figure 3. Data Forwarding

When a node receives a data packet of a session, thenode checks up its Route Table to determine the next action:

1. If no entry associated with this session exists, thenthe data packet is simply dropped.

2. If the associated entry is not active yet, the data packetis queued, and its further action is delayed until the entry

becomes active.3. If the associated entry is active already, the data

packet is transmitted with the corresponding TransmissionPower. Note that forwarding is not needed by this node ifthe Transmission Power is set as 0 (default value as theentry created).

4. Furthermore, if the Membership Flag of the associatedentry is set as IN_GROUP, then the data packet should bepassed to the upper layer and considered as being receivedby a destination.

IV. PERFORMANCE EVALUATIONWe implemented MRMP over NS2 (version 2.31) [10], a

network simulator popularly used in the evaluation ofrouting and multicast protocols, and conducted extensivesimulations. The main objective of the simulations is todemonstrate the performance of MRMP in large-scalemobile ad hoc networks.

A. Simulation Setups and Performance MetricsNS2 provides only omnidirectional antennae in its

current version and the real-world RF radiation may notbehave so ideally with disk-shaped radiation patterns. We


5/6


January 27 - 28, 2011.

465

implemented a user-configurable antenna module toevaluate the capability of MRMP with asymmetriccommunication links (resulted from irregular radiationpatterns). The experimental results of MRMP withomnidirectional and irregular antennae are, respectively,denoted by MRMPO and MRMPI. We also conductedsimulations on MAODV [4], a multicast protocol that wasintended for use in mobile ad hoc networks by the IETFMANET Working Group.

The simulations were studied over networks withdifferent numbers of nodes. Nodes were randomly placed ina 500 500 m2 rectangular area and then kept movingfreely within this area. The Random Walk Mobility Modelwith various maximum velocities was adopted for themobility pattern, where the scenario generator of NS2 wasused to generate topology scenarios. For each topologyscenario, the first 100 nodes were chosen as sources. All ofthe nodes joined the only one multicast group at thebeginning of the simulation and remained as membersthroughout the whole simulation.

Four performance metrics were considered in routingprotocol evaluation. Besides the consideration of energyefficiency, other metrics were also adopted based onpopular routing metrics.

Network lifetime: Measured by the averagenumber of data packets (excluding duplicates) anode can transmit and receive before the first nodeexhausts its energy. This demonstrates theeffectiveness of a routing protocol (and its routingmetric) in the prolongation of network lifetime.

Delivery ratio: The ratio of the number of datapackets successfully delivered to the destinations tothe number of data packets supposed to bereceived. It shows the capability of a routing

protocol to successfully deliver data packets todestinations.

Control overhead: The average number of controlbits transmitted per data bit delivered. The controloverhead includes not only the control messages inthe network layer but also those in the MAC layer.This measures the bit efficiency of a routingprotocol in expending control overhead to deliverydata.

Propagation delay: The average time delayrequired for a data packet to be propagated fromthe source to a destination, including the routediscovery period. It presents the efficiency of aprotocol in data routing.

B. Experimental ResultsFigure. 4 shows the network lifetime with respect to the

network size. The lifetime decreases as the number of nodesincreases, which might violate the intuition that densenetworks might have more alternative routing paths suchthat more energy-efficient routes can be found by MRMP toprolong the network lifetime. Such a phenomenon is due tothe overhearing problem.

Figure.5 shows that the delivery ratios slightly decreaseas the velocity increases. It is because nodes are more likelyto outrun communication ranges. This is the reason whyMAODVP is inferior to MAODVF in terms of data delivery.

In Figure. 6, the control coverheads of both MAODV Fand MAODVP are much heavier than that of MRMPO, andthey increase substantially in general, as the number ofnodes increases.

Figure 4. Network Lifetime

Figure 5. Delivery Ratio

Nodes might produce more control messages in a densenetwork than in a sparse network. The main reason for suchhigh overheads is, nevertheless because of the low deliveryratios.

Figure 6. Control Overhead

Figure. 7 shows that MRMPO is over four times moreefficient than MAODVF and MAODVP in data routing. Thisresult might be contrary to the intuition that MAODV ought


6/6


January 27 - 28, 2011.

466

to be more efficient than MRMP because MAODV attemptsto minimize the number of hops. The reason for thephenomenon is that carrier sensing is done by a node tocompete with its neighbors on the medium usage before thebeginning of a transmission. A large communication rangeimplies a potentially large number of competitors in thenetwork.

Figure 7. Propagation Delay

V. CONCLUSION AND FUTURE WORKThis paper proposes a power-aware routing protocol,

namely the MRMP, in the maximization of the minimumresidual energy of nodes in large-scale mobile ad hocnetworks. MRMP is a source-initiated on-demand multicastprotocol, where no periodic control message is needed in theinformation collection of the network topology andremaining energy. Each multicast tree for a multicast isderived based on the autonomous decision of eachindividual intermediate node and is proved to be

theoretically optimal.For future research, we shall explore the multisource

maximum-residual multicast problem, where multiplesources are considered simultaneously.

REFERENCES

[1] C. Ambuhl, An Optimal Bound for the MST Algorithm to ComputeEnergy Efficient Broadcast Trees in Wireless Networks, Proc. IntlColloquium Automata, Languages, and Programming (ICALP 05),

pp. 1139-1150, 2005.

[2] G. Calinescu, S. Kapoor, A. Olshevsky, and A. Zelikovsky, NetworkLifetime and Power Assignment in Ad-Hoc Wireless Networks,

Proc. European Symp. Algorithms, pp. 114-126, 2003.

[3] J. Chang and L. Tassiulas, Maximum Lifetime Routing in WirelessSensor Networks, IEEE/ACM Trans. Networking, vol. 12, no. 4, pp.609-619, Aug. 2004.

[4] E.M. Royer and C.E. Perkins, Multicast Operation of the Ad HocOn-Demand Distance Vector Routing Protocol, Proc. ACMMobiCom, pp. 207-218, 1999.

[5] C.-C. Chiang, M. Gerla, and L. Zhang, Forwarding Group MulticastProtocol (FGMP) for Multihop, Mobile Wireless Networks,ACM/Baltzer J. Cluster Computing, vol. 1, pp. 187-196, 1998.

[6] J.J. Garcia-Luna-Aceves and E.L. Madruga, The Core - AssistedMesh Protocol, IEEE J. Selected Areas in Comm., vol. 17, no. 8, pp.1380-1994, Aug. 1999.

[7] Z. Genc and O. Ozkasap, EraMobile: Epidemic-Based Reliable andAdaptive Multicast for MANETs, Proc. IEEE Wireless Comm. And

Networking Conf. (WCNC 07), pp. 4395-4400, 2007.

[8] S.-J. Lee, M. Gerla, and C.-C. Chiang, On - Demand MulticastRouting Protocol, Proc. IEEE Wireless Comm. and NetworkingConf. (WCNC 99), pp. 1298-1304, 1999.

[9] L. Ji and M.S. Corson, Differential Destination Multicast-A MANETMulticast Routing Protocol for Small Groups, Proc. IEEEINFOCOM, pp. 1192-1201, 2001.

[10] K. Fall and K. Varadhan, The ns Manual. UC Berkeley and LBL andUSC/ISI and Xerox PARC, 2007.

[11] M. Abolhasan, T.A. Wysocki, and E. Dutkiewicz, A Review ofRouting Protocols for Mobile Ad Hoc Networks, Ad Hoc Networks,vol. 2, pp. 1-22, 2004.

[12] C. de M. Cordeiro, H. Gossain, and D.P. Agrawal, Multicast overWireless Mobile Ad Hoc Networks: Present and Future Directions,IEEE Network, vol. 17, no. 1, pp. 52-59, Jan./Feb. 2003.

[13] S.-J. Lee, W. Su, J. Hsu, M. Gerla, and R. Bagrodia, A PerformanceComparison Study of Ad Hoc Wireless Multicast Protocols, Proc.IEEE INFOCOM, pp. 565-574, 2000.

distributed multicast routing for mobile ad hoc networks

Documents