Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

61 66


Published on

Published in: Technology
  • Be the first to comment

  • Be the first to like this

61 66

  1. 1. ISSN: 2278 – 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012 Energy efficient and Demand based Topology Maintenance for various network traffic conditions Y.Sravana Sandhya V.Sreenathasarma K. V. Nagendra, M.Tech CSE Dept., Asst. Professor, CSE Dept., Asst. Professor, CSE Dept., ASCET, Gudur, ASCET, Gudur, ASCET, Gudur, Andhrapradesh, India. Andhrapradesh, India. Andhrapradesh, India. Abstract—Due to the nodes’ limited resource in the adhoc This method is effective in high data traffic because powernetworks, the scalability is the crucial for network operation. consumption in data transfer dominates the power required toEnergy efficient topology in Ad-hoc networks can be achieved keep nodes awake. We combine the advantages of these twomainly in two different ways. In the first method, networkmaintains a small number of nodes to form a connected techniques to dynamically adjust network topology for variousbackbone and the remaining nodes sleep to conserve energy. This network traffic conditions.method is effective for low traffic networks. Energy efficiency in In this paper, we present a demand based topology (DBT) thatthe second method is achieved by power control technique. This dynamically adjust network topology for various network traffictechnique is effective in high traffic conditions. The first method conditions. We have simulated our proposed protocol DBT byis not effective in high traffic conditions. Similarly, the secondmethod is not effective in low traffic networks. So, in this paper using AODV [8] as routing protocol using network simulatorwe propose a Demand Based Topology (DBT) to reduce the ns2.33 [1] and compared with AODV and AODV with SPAN [3].energy consumption for mobile ad hoc network, by dynamically The simulation studies revealed that the proposed schemeadjusting the topology for various network traffic conditions. We perform better in terms of energy, delay, and delivery ratio. Inhave simulated our proposed protocol DBT by using AODV [8] general network topology is controlled by keeping small numberas routing protocol in network simulator ns2.33 [1] andcompared with AODV and SPAN [3]. The simulation studies of nodes awake as in the first technique. The proposed DBTrevealed that the proposed scheme perform better in terms of keeps more number of nodes along the bulk data transfer path toenergy, delay, and delivery ratio. conserve energy by keeping low link cost as in the second Index Terms—Energy efficient topology, Routing, MANET. technique. The rest of the paper is organized as follows: the next I. INTRODUCTION section provides a brief review of related studies. The third section gives the design details of proposed DBT. Integration Mobile Ad-hoc Networks (MANETs) are self-organizing, issues of DBT with routing protocol is discussed in the forthself-configuring and infrastructure-less multi-hop wireless net- section. Simulation results along with discussions are providedworks, where each node communicates with other nodes in the section 5. The last and final section concludes the paperdirectly or indirectly through intermediate nodes without any with same pointers to future research direction.infrastructure. Such temporary networks can be used in battle-fields, disaster areas, military applications, mining operations II. RELATED WORKand robot data acquisition. Besides these characteristics they We briefly describe various techniques related to our workpresent challenges like limited energy, dynamic topology, low topology control. Different topologies have been proposed inbandwidth and security. The description of the arrangement of the literature to reduce the energy consumption. Thesethe MANETs, called topology, is usually temporary or methods can be classified into centralized controlling anddynamically changed with time. distributed computing methods. Ideally, for mobile ad hoc Energy conserving is one of the challenges because of network a topology should be computed and maintained inlimited battery resource. The techniques which are used to distributed, asynchronous, and localized manner.reduce the initial topology of network to save the energy and Li and Wan [6] described a distributed protocol to constructincrease the lifetime of network, with the preservence of a minimum power topology and developed an algorithmnetwork connectivity, called topology control techniques. which directly find a path whose length is within a constantVarious techniques, in network layer, are proposed in the factor of the shortest path. The length of the path is measuredliterature to conserve energy. These techniques can be in term of energy consumption. This proposed algorithm usedclassified mainly into two categories: by controlling the only local information.number of nodes with the smaller link cost. In the first method A topology based on minimum spanning tree, called local-a small number of nodes awake to maintain the network ized minimum spanning tree (LMST) was proposed by Li etconnectivity and remaining nodes go into sleep state to al. [5]. It is a localized distributed protocol with the followingconserve energy. This method is effective in low traffic properties: (1) the protocol generates a strongly connectedconditions, because the power consumption to keep nodes communication graph; (2) the degree of any node is at mostawake dominates the power consumption in data transfer. Inthe second method, topology is controlled by keeping lessercost links in the network. 61 All Rights Reserved © 2012 IJARCET
  2. 2. ISSN: 2278 – 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012six, and (3) the topology can be made symmetric by removing III. DEMAND BASED TOPOLOGYasymmetric links without impairing connectivity. In this section, we present a demand based energy efficient A simple localized distributed topological control algorithm topology (DBT) for mobile ad hoc network, which(XTC) is proposed by Wattenhpfer et al. Initially each node u dynamically changes the topology according to the networkcomputes a total order (<u) over all its neighbors in the traffic requirements. Initially we compute a small set of nodes,network graph G with respect to decreasing link quality, such which form a connected backbone, while the other nodes areas signal attenuation, and packet arrival rate. Then, each node put off to conserve energy. These connected backbone is usedstart exchanges the neighbour order information among all for routing the packets under low network load. When there isneighbors. At the later point, each node locally selects those a bulk data transfer between a pair of nodes, the topologyneighboring nodes which will form its neighborhood in the dynamically changes along the path between these nodes byresulting topology control graph, based on the previously power control and route optimize technique to minimize theexchanged neighbour order information’s. It covers three main power consumption.advantages: 1) It does not assume the network graph to be a The proposed DBT can be divided into four phases. TheUnit Disk Graph, 2) it works on weighted network graphs, and first phase selects a small set of nodes that constitutes a inde-3) it does not require availability of the node position pendent set of the network. The second phase is responsibleinformation. Authors improved by adding node mobility and for electing more nodes to ensure that the selected nodes formextended XTC for mobile network [7]. a connected backbone. Remaining nodes go to sleep to conserve energy. Active node withdraw process is implement An energy efficient dynamic path is maintained to send data in the third phase to remove redundant nodes in each region.from source to destination for MANET is proposed in Sheu, To improve the performance along the high traffic path we useTu, and Hsu [10]. Due to mobility existing paths may not be the route optimization with power control technique in theenergy efficient. So, each node in a data path dynamically fourth phase. In this technique, we change topologyupdates the path by adjusting its transmission power. Each dynamically to connect more nodes, around the routing path tonode in the networks determines its power for data minimize the total power consumption.transmission and control packets transmission according to thereceived beacon messages from its neighbors. In dynamic path A. Phase I: Independent set formationoptimization technique protocols dynamically select energyefficient path as per the requirement of dynamic topological The first phase selects a minimal set of nodes that constitutechanges in the network [4]. a minimal independent set of a connected backbone of the network. This selection is done in a distributed and localized Another class of topology control protocols based on the k- manner using neighbor information available with the networknearest neighbors graph (k-NNG). In k-NNG every node is layer. Let ni be the total number of nodes surrounding a node iconnected to its k closest neighbors. The Local Information and let nai be the number of additional nodes among theseNo Topology protocol (LINT) [9] try to keep the number of neighbors, which are connected, if node i becomes aneighbors of a node within a low and high threshold centered coordinator to the forward packets. The following heuristic isaround an optimal value. But the optimal value is not used in this phase:characterized. So the estimation of the number of neighbors • Stability factor (denoted by S): Nodes that are relativelycan be inaccurate and connectivity is not guaranteed. The k- more stable as compared to the others in the localities areNeighbors (k-NEIGH) [2] protocol keeps the number of given more preference. The node’s stability is measuredneighbors of a node equal to, or slightly below, a given value as the ratio of number of link failures (fi) and newk. It connects each node with k-closest neighbor, instant of connection established (ci) per unit time to the totalrequire the knowledge of all the neighbors. The number of nodes surrounding that nodei (ni). iTherefore,communication graph that result is made symmetric by c +f stability of a node i is i i . As the values of c and fremoving asymmetric edges, which has k-upper bound ni increase, the stability of the node decreases.neighbors. A characterization of the critical neighbor numbers • Utility factor (denoted by U): Nodes that have higheris discussed by Xue and Kumar. number of neighbors without an active neighbor are Another way of reducing the power consumption is by given more preference. This heuristic is derived from theusing efficient energy path for transmitting packets. These fact that such nodes, if elected, can help a larger numbermethods choose smaller edges in their path to reduce of other nodes, which can then be put to sleep state. Thus, i ni−naitransmission energy. Minimum energy consumption per the utility factor U of a node i is calculate as . nipacket can be achieved by choosing the optimal energy • Energy factor (denoted by E): Nodes that have higherconsumed path from source to destination. However, this amounts of percentage remaining power are given moretechnique does not take the nodes’ energy capacity into preference over others to be elected as active nodes. Thisconsideration. So some nodes may exhaust their power since introduces fairness in the protocol by ensuring properenergy consumption is not fair among the nodes in the rotation in the selection of active nodes. Let E0i denotenetwork. There for the network lift-time decrease . the initial node’s energy and Eti be the amount of energy 62 All Rights Reserved © 2012 IJARCET
  3. 3. ISSN: 2278 – 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012 of a node at time t. So the energy factor Ei of the node • If first phase active nodes may move into a region that i is calculate as E i −E i . 0 E t 0i already has another first phase active node so that the Thus, the above discussion suggests that the coordinator region now has more than one first phase active nodes.selection factor for phase − I can be the sum of all these These active nodes recognize this situation and one offactors them withdraws. ci + fi n n E E • If the withdrawal of a first phase active node may mean Ci = Si + Ui + Ei = + i − ai + 0i − ti (1) E that the second phase active nodes in the locality no ni ni 0i. longer serve their purpose and hence withdraw. Only nodes that do not have an active node in their neighbor- In the above scenarios the respective active nodes withdraw,hood are allowed to participate in the election. Announcement as their need no longer exists. However, when an active nodecontention occurs when multiple nodes discover the lack of an withdraws by virtue of completion of its quota of time it needsactive node, and all decide to become active nodes. We resolve to be awake until another node is elected in the locality.the contention by delaying the announcement with randomized The first three phases of proposed DBT and two phases ofback off delay, which is proportional to the extent to which the SPAN are similar. However DBT have the followingnode satisfies the heuristics. The selected nodes forms an advantages over SPAN.independent set of a connecting backbone of the network. • The proposed DBT uses only 1-hop neighbor informa-Selected active nodes go back to sleep after they have used up a tion to create an energy efficient connected backbonefixed percentage of their power to ensure fairness and allow other network, but the SPAN use 2-hop neighbor information.nodes to become active. • The SPAN consider only energy and utility factor for co- ordinator’s announcement phase, but the proposed DBTB. Phase II: Connecting the Independent Set also consider the stability factor for improving the stabil- Nodes selected in the first phase are not connected. This is ity of the backbone.because there is only one active node in a given locality. In • The SPAN uses same utility factor for all nodes, butthis phase more nodes are elected to ensure that the selected proposed DBT’s utility factors (UA and UB) are dependnodes form a connected network. upon node’s type. Thus, the total number of coordinators All nodes that have two or more active nodes as neighbors, (active nodes) are fewer then the SPAN.which are not connected directly or through one or two active E. Phase IV: Local route customization with Power controlnodes, are eligible to become active in this phase. Preference techniqueis given to the nodes satisfying the following criteria: • Nodes having higher amount of remaining energy. The energy consumption per data packet form source to destination is high when each node uses full transmission • Nodes having higher stability. This can be measured power. This can be reduced by chooses a lower energy cost similar to the one used in the first phase. k Path. The minimum transmission power Pt(d) = ad + c • Nodes having more number of active nodes in the 1-hop is required to send data to a node at a distance d, where neighborhood. 2 < k < 4 and for some2 constants a and c. The receiving The stability and energy factors of this phase is very much h t hr 2 K stsimilar with 1 phase. But the utility factor is depends upon 4 P power Pr = PtGtGr d = t d4 by surface Reflection st model, where ht, Gt, hr, and Gr are respectively antenna heightthe 1 phase’s black active nodes. Let nbi be the number of and gain ofI,J sending and receiving nodes [10]. The stactive nodes of the 1 phase in 1 − hop neighborhood of a P actual power ξ =K t +X, required for sending data from Prnode i. If nodes with high nbi become the coordinators in this a node I to the node J at a distance d, where X represents thephase, fewer coordinators in total may be needed in order to energy consumed by receiving node.make sure every node can talk to a coordinator. Thus a node The minimum required energywith a high nbi should volunteer more quickly than one with for the data transmission can besmaller value. Thus, the coordinator selection factor for calculated as follow: each node in nd2 phase is the sum of all these factors the network has fixed default full ci + f i n n E E transmission power Pt, when a Ci = Si + Ui + Ei = + i − bi + − (2) node I receives control mes-sage 0i ti E ni ni 0i . from node J with power Pr The contention if any is also resolved using the back off Fig. 1. Minimizing the trans- it calculates the distance betweenmechanism like in the first phase. mission power. nodes I and J then node I can find minimum energy Pt(d) re-C. Phase III: Coordinator Withdraw quired for transmitting the data to node J . Let consider the Every active node periodically checks if it should goto sleep nodes B and node C which are in the transmission range of astate or not. The need for a node to be an active may also cease to node A as shown in the Fig. 1. If ξA,B + ξB,C < ξA,C thenexist due to the dynamics of the system. More explicitly, this may sending data packet from node A to the node C viahappen due to one of the following reasons. intermediate node B consume less energy. Our proposed 63 All Rights Reserved © 2012 IJARCET
  4. 4. ISSN: 2278 – 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012DBT uses this power optimization technique locally along the source to itself > as a piggyback with periodic hello messagesrouting path to minimize the energy consumption during the in full transmission path. After receive the hello messagestransmission. from the node I and the node C, along with piggyback Whenever a new node satisfies the above criteria it remains information, node X calculate the link cost ξI,X and ξC,X andawake to participate in the high traffic flow path. Please note checks whether it can participate in the ongoing data transfer.that a new node can come either a sleeping node wakes up The node X can participate in data forwarding, if it reduces thenear high traffic flow path or awake node moves closer to high cost of the path from the node I to the node C. That is, if ξI,Xtraffic flow path. + ξC,X < δI,C then the new node X participate in the routing IV. INTEGRATING DBET WITH by sending route update control message (RUP) to the node I ROUTING PROTOCOL and the node C with route cost δI,C . When the node I and the node C receive (RUP) messages and then update their routing The proposed DBT can be integrated with any routing tables.protocol. In this section, we discusses the process of V. PERFORMANCE EVALUATIONintegration with AODV. In our approach all control packetsand data packets are transmit on low traffic path with full We have evaluated the performance of DBT with AODVtransmission power and data packets on high traffic path with [8] as a routing protocol and compared with AODV andminimum required energy. AODV with SPAN [3] using the network simulator NS2.33.Route discovery: Route discovery uses route cost in place of The simulation parameters are listed in the table I.hop count as route metric. We use the notation δI,J denotes thecost of least cost path from the node I to the node J . TABLE I SIMULATION PARAMETERS When a source node S wants to find a route to a des- Simulation Time 500 sec.tination D, it broadcasts the route request packet (RREQ) to its Number of nodes 20, 40, 60, 80, and 100neighbors. The route request packet contains the least route Max Node Energy 3000 Jcost from source node S, which is initially zero. An Energy Distribution Normal Distribution (0 to 3000 J) 2intermediate node J receiving the route request packet from Area 500×500 m Low traffic load 2×512 bytes/sec.another intermediate node I, it calculates the cost of the path High traffic load 10×512 bytes/sec.form node S to nodes J as δS,I + ξI,J . The node J update its MAC IEEE 802.11routing table if the calculated cost is less than the cost in its Max Tx power 0.2 Wrouting table and forward the route request packet to its Max Rx power 0.09 W Number of power levels 5neighbors with updated cost. In order to avoid another cost Power levels 30 m, 60 m, 90 m, 120 m, and 150 mupdate, node J waits for the time (propositional to the cost to Basic routing protocol AODVδS,J ) before forwarding. Propagation Model Two ray ground propagation model Node motion Random-motion When a destination node D receives first route requestpacket (RREQ), it calculates the route cost and update itsrouting table. It waits for a fixed time interval to receive more Energy consumption: The energy consumed by DBT isroute request packets and find the least cost route among them. same as that of SPAN in low load traffic (see the Fig. 3(a),The node D unicast a route reply packet (RREP) back to its since both the methods keep fewer number of nodes in theneighbor from which it received the least cost route. The backbone and remaining nodes in sleep state to conserve en-neighbor nodes unicast RREP towards the source node S. ergy. However, in high traffic DBT perform better then SPANLocal route customization: As we discussed earlier due to (see the Fig. 3(b)) because of power controlled transmission.the dynamic nature of the network new node may come closer The overhead of keeping more nodes in awake state along theto existing path, which may reduce the existing route cost, if it high traffic, flow path is less than the energy conserved usingparticipates in forwarding the data. power controlled data transmission. From these graphs we can also observe that AODV without using any topology control consumes very high energy. End-to-end delay: The end-to-end delay is the average time between data packets sent out from the sources and received at the destination. The Fig. 3(c) and Fig. 3(d) show the delay with respect to the number of nodes and mobility rate. As the mobility rate increases, the end-to-end delay is Fig. 2. Local path customization. always increases because the network topology changes more frequently. In high network load the DBTs performance is Let consider the example network given in the Fig. 2(a) better then SPAN and AODV. This is because the lowwith the existing path cost from the node I to the node J is 9 transmission power implies low queuing delay and reducedunits. If a node is in data transmission path, it sends the interference. Similarly, the increase in delay become more< Source address, Destination address, Route cost from 64 All Rights Reserved © 2012 IJARCET
  5. 5. ISSN: 2278 – 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012 (a) Low network load (b) High network load (c) High network load Consumed energy v/s Simulation time. Delay v/s number of nodes for mobility=2. (d) High network load (e) Low network load (f) High network load Delay v/s mobility. Packet delivery ratio v/s mobility. Fig. 3. Simulation Results.significant when the network density increases because more [3] Benjie Chen, Kyle Jamieson, Hari Balakrishnan, and Robertnumber of nodes creates more interfaces in the routing. Morris.Span: An energy-efficient coordination algorithm for Packet delivery ratio: Packet delivery ratio is the ratio of topology mainte- nance in ad hoc wireless networks. ACM Wireless Networks, 8:85–96, September 2001.the data packets received at the destination to the data packets [4] H. P. Gupta and S. V. Rao. Pclr: Power control-based locallysent out from the sources. The Fig. 3(e) and Fig. 3(f) show the customize routing for Manet. In Proc. IEEE Internationaloverall delivery ratio with respect to the mobility rate. As the Conference on RF and Signal Processing Systems, pages 632– 637, 2010.mobility rate increases, the delivery ratio always decrease. In [5] N. Li, J.C. Hou, and L. Sha. Design and analysis of an mst-low network load, there is not much difference between based topology control algorithm. In INFOCOM 2003. Twenty-proposed DBT and SPAN. But if the network load is high Second Annual Joint Conference of the IEEE Computer and Communications. IEEE Societies, pages 1702 – 1712, mar.DBTs performance drops slower than SPAN. This is because, 2003. Xiang-Yang Li and Peng-Jun Wan. Constructingin SPAN the probability of packet loss increases with increase minimum energy mobile wireless networks. SIGMOBILEin traffic due to collisions and interfering. But DBT over Mob. Comput. Commun. Rev.,5(4):55–672001. [6] Xiang-Yang Li and Peng-Jun Wan. Constructing minimumcomes the problem by route optimize phase. energy mobile wireless networks. SIGMOBILE Mob. Comput. Commun. Rev.,5(4):5567, 2001. VI. CONCLUSIONS [7] Naghshegar, A. Dana, A. Darehshoorzadeh, and K. Karimpoor.Topology control scheme in manets for aodv [1] In this paper, we proposed a demand based energy routing. In Information and Communication efficient topology that dynamically adjusts its Technologies: From Theory to Applications, 2008. ICTTA 2008. 3rd International Conference on, pages 1 –6, 7-11 2008. topology for various network traffic conditions. [8] Charles E. Perkins and Elizabeth M. Royer. Ad-hoc on-demand We have simulated our proposed protocol DBT distance vector routing. In WMCSA ’99: Proceedings of the and compared with AODV and AODV with Second IEEE Workshop on Mobile Computer Systems and Applications, pages 90– 100, Washington, DC, USA, 1999. SPAN. The simulation studies revealed that the IEEE Computer Society. proposed scheme perform better in terms of [9] Ram Ramanathan and Regina Hain. Topology control of energy, delay, and delivery ratio. multihop wireless networks using transmit power adjustment. In INFOCOM, pages 404–413, 2000. REFERENCES [10] Jang-Ping Sheu, Shin-Chih Tu, and Chi-Hung Hsu. Location- [1] The network simulator - ns-2.33. free topol- ogy control protocol in wireless ad hoc networks. [2] Douglas M. Blough, Mauro Leoncini, Giovanni Resta, and Comput. Commun.,31(14):341032008. Paolo Santi. The k-neighbors approach to interference bounded and symmetric topology control in ad hoc networks. IEEE Transactions on Mobile Computing, 5(9):1267–1282, 2006. 65 All Rights Reserved © 2012 IJARCET
  6. 6. ISSN: 2278 – 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012Author Profile:Y.Sravana Sandhya MTECH (SE) from ASCETGudur,Andhrapradesh,India,Area of interest incomputer networking and software engineering.V.sreenathasarma Asst.Professor (CSE) from ASCETGudur, Andhrapradesh, India, Area of interest incomputer networking and Datamining and warehousing.. K.V.Nagendra Asst.Professor (CSE) from ASCETGudur, Andhrapradesh, India, Area of interest incomputer networking and Data mining and warehousing. 66 All Rights Reserved © 2012 IJARCET