Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more
Download
Standard view
Full view
of .
Save to My Library
Look up keyword
Like this
1Activity
0 of .
Results for:
No results containing your search query
P. 1
Enhanced Dynamic Source Routing Protocol using On Demand Passive Clustering

Enhanced Dynamic Source Routing Protocol using On Demand Passive Clustering

Ratings: (0)|Views: 50 |Likes:
Published by ijcsis
In this paper an enhancement technique for Dynamic Source Routing (DSR) protocol using on demand passive clustering is proposed. DSR is a popular reactive protocol in Mobile Adhoc Networks (MANETs) which uses flooding for route discovery and route maintenance. Flooding can be easily restricted using clustering. Clustering restricts the set of forwarding nodes during flooding which in turn reduces the energy, cost and traffic overhead during routing in dynamic traffic and dynamic topology environment of MANET. Active clustering mechanisms require periodic refresh of neighborhood information and tend to introduce a large amount of communication maintenance overhead. This can be overcome using passive clustering which executes only when the user has data to transmit. Passive clustering dynamically partitions the network into clusters interconnected by gateways. Existing Passive Clustering techniques exploit the data packets for cluster formation and thus reduce the periodic exchange of explicit control packets. We have proposed an enhanced passive clustering technique which makes use of the control packets used for route discovery in DSR to establish a cluster. We have used the Lowest ID technique and first declaration wins technique to form clusters. We have analyzed the performance of this technique to reduce flooding by using single cluster head called primary cluster head. Then an analysis has been carried out with two cluster heads called primary and secondary cluster heads. Simulation results on DSR have shown that this enhanced passive clustering technique can reduce redundant flooding, with negligible overhead; - thus making DSR more efficient.
In this paper an enhancement technique for Dynamic Source Routing (DSR) protocol using on demand passive clustering is proposed. DSR is a popular reactive protocol in Mobile Adhoc Networks (MANETs) which uses flooding for route discovery and route maintenance. Flooding can be easily restricted using clustering. Clustering restricts the set of forwarding nodes during flooding which in turn reduces the energy, cost and traffic overhead during routing in dynamic traffic and dynamic topology environment of MANET. Active clustering mechanisms require periodic refresh of neighborhood information and tend to introduce a large amount of communication maintenance overhead. This can be overcome using passive clustering which executes only when the user has data to transmit. Passive clustering dynamically partitions the network into clusters interconnected by gateways. Existing Passive Clustering techniques exploit the data packets for cluster formation and thus reduce the periodic exchange of explicit control packets. We have proposed an enhanced passive clustering technique which makes use of the control packets used for route discovery in DSR to establish a cluster. We have used the Lowest ID technique and first declaration wins technique to form clusters. We have analyzed the performance of this technique to reduce flooding by using single cluster head called primary cluster head. Then an analysis has been carried out with two cluster heads called primary and secondary cluster heads. Simulation results on DSR have shown that this enhanced passive clustering technique can reduce redundant flooding, with negligible overhead; - thus making DSR more efficient.

More info:

Published by: ijcsis on Sep 05, 2010
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

09/05/2010

pdf

text

original

 
Enhanced Dynamic Source Routing Protocol usingOn Demand Passive Clustering
Shobha.K.R and Dr.K.Rajanikanth
M.S.Ramaiah Institute of TechnologyBangalore, India.shobha_shankar@yahoo.comprincipal@msrit.edu 
 Abstract
 — 
 
In this paper an enhancement technique for DynamicSource Routing (DSR) protocol using on demand passiveclustering is proposed. DSR is a popular reactive protocol inMobile Adhoc Networks (MANETs) which uses flooding forroute discovery and route maintenance. Flooding can be easilyrestricted using clustering. Clustering restricts the set of forwarding nodes during flooding which in turn reduces theenergy, cost and traffic overhead during routing in dynamictraffic and dynamic topology environment of MANET. Activeclustering mechanisms require periodic refresh of neighborhoodinformation and tend to introduce a large amount of communication maintenance overhead. This can be overcomeusing passive clustering which executes only when the user hasdata to transmit. Passive clustering dynamically partitions thenetwork into clusters interconnected by gateways. ExistingPassive Clustering techniques exploit the data packets for clusterformation and thus reduce the periodic exchange of explicitcontrol packets. We have proposed an enhanced passiveclustering technique which makes use of the control packets usedfor route discovery in DSR to establish a cluster. We have usedthe Lowest ID technique and first declaration wins technique toform clusters. We have analyzed the performance of thistechnique to reduce flooding by using single cluster head calledprimary cluster head. Then an analysis has been carried out withtwo cluster heads called primary and secondary cluster heads.Simulation results on DSR have shown that this enhanced passiveclustering technique
can reduce redundant flooding
, withnegligible overhead; - thus making DSR more efficient.
 Keywords- MANETs; DSR; Flooding; Clustering; Passive clustering; Primary cluster head; Secondary cluster head;
I.
 
I
NTRODUCTION
Mobile Adhoc Networks (MANETs) have recently been thesubject of active research because of their unique advantages.MANETs are self-creating, self-organizing and self-administrating networks and do not require deployment of anykind of infrastructure.
They offer special benefits and
versatility for wide range of applications in military (e.g.,
 battlefields, sensor networks etc.), commercial (e.g., distributed
mobile computing, disaster discovery systems, etc.), andeducational environments (e.g., conferences, conventions, etc.),where fixed infrastructure cannot be easily acquired. With theabsence of pre-established infrastructure (e.g., no router, noaccess point, etc.), two nodes communicate with one another ina peer-to-peer fashion. Two nodes communicate directly if theyare within transmission range of each other. Otherwise, nodesmust communicate via a multihop route. To find such a multi-hop route, MANETs commonly employ on demand routing
algorithms that use flooding or broadcast messages. Many ad
hoc routing protocols [12] [13] [26] [27], multicast schemes[25], or service discovery programs depend on massive
flooding. In flooding, a node transmits a message to all of its
neighbors. The neighbors in turn relay to their neighbors andso on until the message has been propagated to the entire
network. In this paper, we will refer to such flooding as bli
ndflooding. As one can easily see, the performance of blind
ooding is closely related to the average number of neighbors(neighbor degree) in the CSMA/CA network. As the neighbor
degree gets higher, blind flooding suffers from the increase of 
(1) redu
ndant and superfluous packets, (2) probability of 
collision, and (3) congestion of wireless medium [1].
Performance of blind flooding is severely impaired especially
in large and dense networks [2]. When topology orneighborhood information is available, only subsets of 
neighbors are required to participate in flooding to guaranteethe complete flooding. We call such flooding
as efficient
ooding. The characteristics of MANETs (e.g., node mobility,the limited bandwidth and resource), however, make theperiodic collection of topology information difficult and costly(in terms of overhead). For this reason many on-demand adhocrouting schemes and service discovery protocols simply useblind
ooding [12] [25]. In contrast with on-demand routingmethods, the proactive adhoc routing schemes by virtue of periodic route table exchange, can gather topologicalinformation without much extra overhead. Thus, the leadingMANET proactive adhoc routing schemes use routeaggregation methods to forward routing packets through only asubset of the neighbors [26] [27].In this paper, we focus on on-demand routing protocols andpropose a
mechanism for efficient flooding based on passive
clustering. Existing passive clustering scheme [10] requiresneither the deployment of GPS like systems nor explicitperiodic control messages to identify the subset of forwardingneighbors. This scheme makes the following contributions
compared with previous efficient flooding schemes (such as
multipoint relay, neighbor coverage, etc): (1) it does not needany periodic messages. Instead, it exploits existing datapackets by attaching few more extra fields; (2) it is veryresource-efficient regardless of the degree of neighbor nodes or
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 5, August 201027http://sites.google.com/site/ijcsis/ISSN 1947-5500
 
the size of network. To our knowledge, passive clustering isthe only scheme that provides scalability and practicality forchoosing the minimal number of forwarding nodes in thepresence of dynamic topology changes; (3) it does notintroduce any startup latency; (4) it saves energy if there is notraffic; (5) it easily adapts to topology and available resourcechanges.In this paper we propose an enhanced passive clusteringscheme which uses route request packets (RREQ) instead of data packets for maintaining cluster. This technique reduces theflooding efficiently and also reduces the overhead in each andevery data packet transmitted. We have simulated and testedthis algorithm using DSR routing protocol.The remainder of the article is organised as follows: Insection 2 we discuss the various methods available forachieving efficient flooding. Section 3 gives an explanationabout the algorithms we are proposing for achieving efficientflooding .Section 4 discusses the simulation parameters andresults .The main conclusions for this paper are summarized insection 5.II.
 
R
ELATED WORK
 Several papers [1] [6] [7] [8] have addressed the limitations
of blind flooding and have prop
osed solutions to provideefficient
ooding.
However, because of the problem of finding
a subset of dominant forwarding nodes in MANETs, all the
work about efficient flooding has been directed to the
development of efficient heuristics that select a sub-optimaldominant set with low forwarding overhead.In [1] [6], the authors propose several heuristics to reduce
rebroadcasts. More specifically, upon receiving a flood packet,
a node decides whether to relay it or not based on one of thefollowing heuristics: (1) rebroadcast with given probability;(2) rebroadcast if the number of received duplicate packets isless than a threshold; (3) distance-based scheme where therelative distance between hosts determines the rebroadcastdecision; (4) location-based scheme where the decision is basedon pre-acquired neighbor location information; (5) cluster-based scheme where only precomputed cluster heads andgateways rebroadcast. Our approach, passive clustering, differsfrom the above schemes in that it provides a more systematicmethod based on locally collected information (e.g., neighborinformation, cluster states, etc.). Each node participates in
ooding based on its role or state in the dynamicallyconstructed cluster architecture instead of depending on localheuristics or on pre-computed clusters.
Another approach to efficient flooding is to exploit
topological information [6] [7] [8] [24] [19]. In the absence of preexisting infrastructure, all the above schemes use a periodichello message exchange method to collect topologicalinformation. Our approach does not require periodic controlmessages. Rather, it exploits on-going Route REQuest(RREQ) packets to exchange cluster-related information. Theauthors of [8] suggest two schemes called self-pruning anddominant pruning. Self pruning is similar to the neighbor-coverage scheme in [6]. With self-pruning scheme, eachforwarding node piggybacks the list of its neighbors onoutgoing packet. A node rebroadcasts (becomes a forwardingnode) only when it has neighbors that are not covered by itsforwarding nodes. While the self-pruning heuristic utilizesinformation of directly connected neighbors only, thedominant-pruning heuristic extends the propagation of neighbor information two-hop away. The dominant pruningscheme is actually similar to Multipoint Relay scheme (MPR)[7]. In Multipoint Relay scheme, a node periodicallyexchanges the list of adjacent nodes with its neighbors so thateach node can collect the information of two-hop awayneighbors. Each node, based on the gathered information,selects the minimal subset of forwarding neighbors, whichcover all nodes within two-hops. Each sender piggybacks itschosen forwarding nodes (i.e. multipoint relay nodes) on theoutgoing broadcast packet.
Figure 1. The collision rate of broadcast
Along the same lines, several other schemes have proposedthe selection of a dominant set based on topology [18] [21] [22][23]. All of these schemes, however, again depend on periodichello messages to collect topological information.The extra hello messages, however, consume resources anddrop the network throughput in MANETs [14]. The extratraffic brings about congestion and collision as geographicdensity increases [1]. Fig 1[26] depicts the collision probabilityof hello messages in a single hop and a two hop network as the
number of neighbor’s increases. This result clearly shows that
the neighbor degree causes the broadcast collision probabilityto increase (note, the collision probability is more than 0.1 withmore than 15 neighbors). Moreover, the hidden terminal andexposed terminal problems aggravate collisions in the two hopnetwork. Note that Fig 1 assumes no data traffic - only hellomessages.With user-data packets, the collision probability of hellomessages will dramatically increase. Thus, it will be hard tocollect complete neighbor topology information using hellomessages. As a consequence, the aforementioned schemes(e.g., neighbor coverage, MPR, etc.) are not scalable to offeredload and number of neighbors.Lastly, we consider clustering. Clustering can be describedas grouping of nodes. A representative of each group (cluster)is dynamically elected to the role of cluster head based on somecriterion (e.g., lowest ID). Nodes within one hop of a clusterhead become associated to its cluster. A node belonging to twoor more clusters at the same time is called a gateway. Other
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 5, August 201028http://sites.google.com/site/ijcsis/ISSN 1947-5500
 
members are called ordinary nodes. Various distributed
Figure 2. . An example of efficient flooding with clustering. Only clusterheads and gateways rebroadcast
computation techniques can be used to dynamically createclusters. In the active clustering lowest ID technique [15] eachnode attempts to become cluster head by broadcasting its ID toneighbors. It will give up only if it hears from a lower IDneighbor. Based on the above definition, any two nodes in acluster are at most 2 hops away [9]. With this clusteringscheme, the dominant forwarding nodes are the cluster headsand the gateways, as shown in Figure 2.Clustering in ad hoc networks has been extensively studiedfor hierarchical routing schemes [9] [5] [3], the master electionalgorithms [4], power control [3], reliable broadcast [20], loadaware schemes [17], efficient broadcast [16] and efficientflooding [10]. Some clustering schemes are based on thecomplete knowledge of neighbors. However, the completeknowledge of neighbor information in adhoc networks is hardto collect and introduces substantial control overhead caused byperiodic exchange of hello messages. Passive clustering [10] is
an “on demand” protocol. It constructs and maintains the
cluster architecture only when there are on-going data packets
that piggyback “cluster related information”. Each node
collects neighbor information through promiscuous packetreceptions. Passive clustering, therefore, eliminates setuplatency and major control overhead of clustering protocols.Passive clustering has two innovative mechanisms for thecluster formation: First Declaration Wins rule and GatewaySelection Heuristic. With the First Declaration Wins rule, a
node that first claims to be a cluster head “rules” the rest of 
nodes in its clustered area (radio coverage). There is nowaiting period (to make sure all the neighbors have beenchecked) unlike for all the weight-driven clusteringmechanisms [3] [5]. Also, the Gateway Selection Heuristicprovides a procedure to elect the minimal number of gateways(including distributed gateways) required to maintain theconnectivity in a distributed manner.Passive clustering maintains clusters using implicit timeout.A node assumes that the nodes it had previously heard from asdead or out of its locality if they have not sent any data withintimeout duration. With a reasonable offered load, a node caneasily keep track of dynamic topology changes by virtue of thistimeout.Lastly, the existing passive clustering scheme usesflooding initially till the cluster is formed and each data packetcarries cluster formation overhead fields even though thecluster has not changed. These limitations motivated ourinvestigation of a new cluster formation protocol calledenhanced passive clustering. While retaining the advantages of passive clustering, our scheme eliminates much of the controloverhead.III.
 
E
NHANCED PASSIVE CLUSTERING
 
 A.
 
Overveiw of Enhanced passive clustering
The proposed protocol called enhanced passive clustering isvery much similar to passive clustering algorithm except that ituses RREQ packets for formation of clusters and theirmaintenance. RREQ packets are generated only when thetopology of the network changes and hence is more suitable forcluster formation and maintenance in a rapidly changingnetwork. RREQ packet is transmitted by the node having datato transmit but has no path to forward it. If data packets areused for cluster formation and maintenance then every datapacket carries the overhead of cluster related information thusincreasing the overhead transmitted in the network. Ourscheme helps in reducing flooding and redundant overheadtransmission effectively.
 B.
 
Construction and maintainance
A node can be in different states during the clusteringprocess, namely: INITIAL, CLUSTER HEAD (CH),ORDINARY NODE, GATEWAY (GW), CH READY, GWREADY and DISTRIBUTED GW.When a node joins the network, it sets its cluster state toINITIAL. Moreover, the state of a floating node (a node doesnot belong to a cluster yet) also is set to INITIAL. Becauseenhanced passive clustering exploits RREQ packets, theimplementation of this clustering resides between layer 3 and 4.An additional field in the header (the cluster information field)is carried by RREQ packet. This field contains the followingentries:
 
Node ID: The IP address of the sender node. Not to beconfused with the source address of the IP packet.
 
State of node: The cluster state of the sender node.
 
Two cluster heads addresses: If a sender node is agateway, then there is another field with the two IPaddresses of the cluster heads (CHs) which arereachable from the gateway.Below, we provide a summary description of the enhancedpassive clustering algorithm.
1)
 
The packet handling:
Upon sending a RREQ packet, each node piggybackscluster related information in the cluster info field. Upon apromiscuous packet reception, each node extracts cluster-related information of neighbors and updates neighborinformation table.
2)
 
 A cluster head declaration:
A node in INITIAL state changes its state to CH READY(a candidate cluster head) when a RREQ packet arrives fromanother node that is not a cluster head. With the next outgoingRREQ packet, a CH READY node can declare itself as aCLUSTER HEAD.
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 5, August 201029http://sites.google.com/site/ijcsis/ISSN 1947-5500

You're Reading a Free Preview

Download
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->