a review on current routing protocols in wireless ad hoc netwrks

8/6/2019 A Review on Current Routing Protocols in Wireless Ad Hoc Netwrks

1/10

AbstractAn ad hoc mobile network is a collection of mobile nodes that are dynamically and arbitrarily located in such a manner that th e interconnectionsbetween nodes are capable of changing on a continual basis. In order to facilitate communicationwithin the network, a routing protocol is usedto discover routes between nodes. The primary goal of such an ad hoc network routing protocol is correct and efficient route establishmentbetween a pair of nodes so that messages may be delivered in a timely manner. Route construction should be done with a minimum of overheadand bandwidth consumption. This article examines routing protocols fo r ad hoc networks and evaluates these protocols based on a given set ofparameters. The: article provides an overview of eight differcnt protocols by presenting their characteristicsand functionality,and then provides acomparison and discussion of their respective merits and drawbacks.A Review o f Current Routing ProtocolsAd Hoc Mobile Wireless Networks for

ELIZABETHM . ROYER, UNIVERSITY O F C A L I F O R N I A ,AN T AA R B A R ACHAI-KEONGTOH,G E O R G I ANSTITUTE OF TECHNOLOGY

ince their emergence in the1970s,wireless networks have become increasingly popular inthe computing industry. This is particularly true within thepast decade, which has seen wireless networks being adaptedto enable mobility. There are currently two variations ofmobile wireless networks. The first is known as the infrastruc-tured network (i.e., a network with fixed and wired gateways).The bridges for these networks are known as base stations. Amobile unit within these networks connects to, and communi-cates with, the nearest base station that is within its communi-cation radius. As the mobile travels out of range of one basestation and into the range of another, a handoff occurs fromthe old base station to the new, and the mobile is able to con-tinue communication seamlessly throughout the network. Typ-ical applications of this type of network include office wirelesslocal area networks (WLANs).The second type of mobile wireless network is the infras-tructureless mobile network, commonly known as an ad hocnetwork. Infrastructureless networks have no fixed routers; allnodes are capable of movement and can be connected dynami-cally in an arbitrary manner. Nodes of these networks functionas routers which discover and maintain routes to other nodesin the network. Example applications of ad hoc networks areemergency search-and-rescue operations, meetings or conven-tions in which persons wish to quickly share information, anddata acquisition operations in inhospitable terrain.This article examines routing protocols designed for thesead hoc networks by first describing the operation of each ofthe protocols and then comparing their various characteris-tics. The remainder of the article is organized as follows.The next section presents a discussion of two subdivisions ofad hoc routing protocols. Another section discusses currenttable-driven protocols, while a la ter section describes thoseprotocols which are classified as on-demand. The articlethen presents qualitative comparisons of table-driven proto-cols, followed by demand-driven protocols, and finally a gen-eral comparison of table-driven and on-demand protocols.Applications and challenges facing ad hoc mobile wirelessnetworks are discussed; and finally, the last section con-cludes the article.

Existing Ad Hoc Routing ProtocolsSince the advent of Defense Advanced Research Proje ctsAgency (DAR PA) packet radio networks in the early 1970s[ l ] , umerous protocols have been developed for ad hocmobile networks. Such protocols must deal with th e typicallimitations of these networks, which include high powerconsumption, low bandwidth, and high err or rates . Asshown in Fig. 1, these rout ing protocols may generally becategorized as:Table-drivenSource-initiated (demand-driven)

Solid lines in this figure represent direct descendants, whiledotted lines depict logical descendants. Despite being designedfor the same type of underlying network, the characteristics ofeach of these protocols are quite distinct. The following sec-tions describe the protocols and categorize them according totheir characteristics.Table-Driven Routina ProtocoIsJTable-driven routing protocols attempt to maintain consistent,up-to-date routing information from each node t o every othernode in the network. These protocols require each node tomaintain one or more tables to s tore routing information, andthey respond to changes in network topology by propagatingupdates throughout the network in order to maintain a consis-tent network view. The areas in which they differ are thenumber of necessary routing-related tables and the methodsby which changes in network structure are broadcast. T he fol-

lowing sections discuss some of the existing table-driven adhoc routing protocols.Destination-Sequen ced Distance-V ector Routing - The Des-tination-Sequenced Distance-V ector Routing protocol(DSDV) described in [2] is a table-driven algorithm based onthe classical Bellman-Ford ro uting mechanism [ 3 ] .Theimprovements made to the Bellman-Ford algorithm includefreedom from loops in routing tables.Every mobile node in the network maintains a routingtable in which all of the possible destinations within the net-

46 1070-9916/99/$10.00 0 999 IEEE IEEE Personal Communications * April 1999


2/10

work and the number of hops to each destination are record-ed. Each entry is marked with a sequence number assigned bythe destination node. The sequence numbers enable themobile nodes to distinguish stale routes from new ones, there-by avoiding the forma tion of routing loops. Routing t ableupdates a re periodically transmitted throughout the networkin order to maintain table consistency. To help alleviate thepotentially large amount of network traffic that such updatescan generate, route updates can employ two possible types ofpackets. The first is known as afull dump. This type of packetcarries all available routing information and can require mul-tiple network protocol data units (NPDUs). During periods ofoccasional movement, these packets ar e transmitted infre-quently. Smaller incremental packets ar e used to relay onlythat information which has changed since the last full dump.Each of these broadcasts should fit into a standard-sizeNPDU, thereby decreasing the amount of traffic generated.The mobile nodes maintain an additional table where theystore the data sent in the incremental routing informationpackets.New route broadcasts contain the address of t he destina-tion, the number of hops to reach the destination, thesequence number of the information received regarding thedestination, as well as a new sequence number unique to th ebroadcast [ 2 ] . The route labeled with the most recentsequence number is always used. In th e event that twoupdates have the same sequence number, the r oute with thesmaller metric is used in order to optimize (shorte n) thepath. Mobiles also keep track of the settling time of route s,or the weighted average time that routes to a destination willfluctuate before t he route with th e best metric is received(see [ 2 ] ) .By delaying the broadcast of a routing update bythe length of the settling time, mobiles can reduce networktraffic and optimize routes by eliminating those broadcaststhat would occur if a better route was discovered in the verynear future.Cl u s terh ea d G a t ew a y S wi t ch Ro u t i ng - The ClusterheadGateway Switch Routing (CGSR) protocol differs from theprevious protocol in the type of addressing and networkorganization scheme employed. Ins tead of a flat network,CGSR is a clustered multihop mobile wireless network withseveral heuristic routing schemes [4]. The authors state thatby having a clus ter head controlling a grou p of ad h ocnodes, a framework for code separation (among clusters),channel access, routing, and bandwidth al location can beachieved. A cluster head selection algorithm is utilized toelect a node as the cluster head using a distributed algo-

H Figure2. CGS R: routingfrom node 1 to node 8.

W Figure 1. Categorization of ad hoc routingprotocols.

rithm within the cluster. The disadvantage of having a clus-ter head scheme is that frequent cluster head changes canadversely affect routing protocol performance since nodesare busy in cluster head selection rather than packet relay-ing. Hence, ins tead of invoking cluster hea d reselectionevery time the cluster membership changes, a Least ClusterChange (LCC) clustering algorithm is introduced. UsingLCC, cluster heads only change when two cluster headscome into contact, or when a node moves out of contact ofall other cluster heads.CGSR uses DSDV as the underlying routing scheme, andhence has much of the same overhead as DSDV. However, itmodifies DSDV by using a hierarchical cluster-head-to-gate-way routing approach to route traffic from source to destina-tion. Gateway nodes are nodes that are within communicationrange of two or more cluster heads. A packet sent by a nodeis first routed to its cluster head, and then the packet is rout-ed from the cluster head to a gateway to ano ther cluster head,and so on until the cluster head of the destination node isreached. The packet is then t ransmitted to th e destination.

Figure 2 illustrates an example of this routing scheme. Usingthis method, each node must keep a cluster member tablewhere it stores the destination cluster head for each mobilenode in the network. These cluster member tables are broad-cast by each node periodically using the DSDV algorithm.Nodes update their cluster member tables on reception ofsuch a table from a neighbor.In addit ion to the cluster member table, each nodemust also maintain a routing table which is used to deter-mine the next ho p in order to reach t he destination. Onreceiving a packet, a node will consult its cluster membertable and routing table to determine the nearest clusterhead along the route to the destination. Next, the nodewill check its routing table to determine the next hop usedto reach t he selected cluster head. It then transmits thepacket to this node.The Wireless Routing Protocol - The Wireless Routing Pro-tocol (WRP) described in [ 5 ] is a table-based protocol withthe goal of maintaining routing information among all nodesin the network. Each node in the network is responsible formaintaining four tables:Distance table

Routing tableLink-cost tableMessage retransmission list (MRL) tableEach entry of the MRL contains the sequence number of theupdate message, a retransmission counter, an acknowledg-ment-required flag vector with one entry per neighbor, and alist of updates sent in the update message. The MRL recordswhich updates in an update message need t o be retransmit-ted and which neighbors should acknowledge the retransmis-sion [ 5] .Mobiles inform each other of link changes through the use

IEEE Personal Communications April 1999 47


3/10

of update messages. An update message is sent only betweenneighboring nodes and contains a list of updates (the destina-tion, the distance to the destination, and the predecessor ofthe destina tion), as well as a list of responses indicatingwhich mobiles should acknowledge (ACK) the update.Mobiles send update messages after processing updates fromneighbors or detecting a change in a link to a neighbor. Inthe event of the loss of a link between two nodes, the nodessend update messages to their neighbors. The neighbors thenmodify their distance table entries and check for new possiblepaths through o ther nodes. Any new paths are relayed backto t he original nodes so that they can update their tablesaccordingly.Nodes learn of the existence of their neighbors from thereceipt of acknowledgments and other messages. If a node isnot sending messages, it must send a hello message within aspecified time period to ensure connectivity. Otherwise, thelack of messages from the node indicates th e failure of thatlink; this may cause a false alarm. When a mobile receives ahello message from a new node, that new node is added to themobiles routing table, and the mobile sends the new node acopy of its routing table information.Part of the novelty of WRP stems from the way in whichit achieves locip freedom. In WRP, routing nodes communi-cate th e distance and second-to-last hop information foreach destination in the wireless networks. WRP belongs tothe class of path-finding algorithms with an important excep-tion. It avoids th e count-to-infinity problem [6] by forcingeach node to perform consistency checks of predecessorinformation re port ed by all its neighbors. This ultimately(although not instantaneously) eliminates looping situationsand provides faster route convergence when a link failureevent occurs.

Source-Initiated On-Demand RoutingA different approach from table-driven routing is source-initi-ated on-demand routing. This type of routing creates routesonly when desired by the source node. When a node requiresa route t o a destination, it initiates a route discovery processwithin the network. This process is completed once a route isfound or all possible route permutations have been examined.Once a route has been established, it is maintained by a routemaintenance procedure until eit her the destination becomesinaccessible along every path from the source or until theroute is no longer desired.Ad Hoc On-D ema nd Distance Vector Routing - The Ad HocOn-Demand Distance Vector (AODV) routing protocoldescribed in [ 7 ]builds on t he DSDV algorithm previouslydescribed. AODV is an improvement on DSDV because ittypically minimizes the number of required broadcasts by cre-ating routes on a demand basis, as opposed to maintaining acomplete list of routes as in the DSDV algorithm. The authorsof AODV classify it as apure on-demand route acquisition sys-tem, since nodes that are not on a selected path do not main-tain routing information or participate in routing tableexchanges [ 7 ] .When a source node desires to send a message to somedestination node and does not already have a valid route tothat destination, it initiates a path discovery process t o locatethe other node. It broadcasts a route request ( RREQ ) packetto its neighbors, which then forward the request to theirneighbors, and so on, until either the destination or an inter-mediate node with a fresh enough route to t he destinationis located. Figure 3a illustrates the propagation of the broad-cast RREQs across the network. AODV utilizes destinationsequence numbers to ensure all routes are loop-free and con-

tain the most recent route information. Each node maintainsits own seque nce number, as well as a broadcast ID. Thebroadcast ID is incremented for every RREQ the node initi-ates, and together with the nodes IP address, uniquely identi-fies an RREQ. Along with its own sequence number and thebroadcast ID, the source node includes in the R RE Q themost recent sequence number it has for the destination. Inter-mediate node s can reply to the RR EQ only if they have aroute to the destination whose corresponding destinationsequence number is greater than or equal to that contained inthe RREQ.During the process of forwarding the RREQ, intermediatenodes record in their route tables the address of the neighborfrom which the first copy of the broadcast packet is received,thereby establishing a reverse path. If additional copies of thesame RREQ are later received, these packets are discarded.Once the RREQ reaches the destination or an intermediatenode with a fresh enough rou te, the destinationiintermediatenode responds by unicasting a route reply (RREP) packetback to the neighbor from which it first received the RR EQ(Fig. 3b). As the RREP is routed back along the reverse path,nodes along this path set up forward route entries in theirroute tables which point to the node from which the RREPcame. These forward route entries indicate the active forwardroute. Associated with each route entry is a route timer whichwill cause the deletion of the entry if it is not used within thespecified lifetime. Because the RREP is forwarded along thepath established by the RR EQ, AODV only supports the useof symmetric links.Route s are maintained as follows. If a source node moves,it is able to reinitiate the route discovery protocol to find anew route to the destination. If a node along the route moves,its upstream neighbor notices the move and propagates a linkfailure no tification message (an RREP with infinite metric) toeach of its active upstream neighbors to inform th em of theerasure of that part of the route [ 7 ] .These nodes in turn

Destination

Source

U(a) Propagation of th e RREQ/m - p Destination

Source

Ly(b) Path of th e RREP to the source

I igure 3.AODVroute discovety.

48 IEEE Personal Communications April 1999


4/10

IU Figure4. Creation of the route record in DSR.propagate the link ailure notification to thei r upstream neigh-bors, and so on until the source node is reached. The sourcenode may then choose to reinit iate rou te discovery for thatdestination if a route is still desired.An addit ional aspect of the p rotocol is the use of hellomessages, periodic local broadcasts by a node to inform eachmobile node of other nodes in its neighborhood. Hello mes-sages can be used t o maintain t he local connectivity of anode. However, th e use of hello messages is not required.Nodes listen for retransmission of data packets to ensure thatthe next hop is still within reach. If such a retransmission isnot heard, the node may use any one of a number of tech-niques, including the reception of hello messages, to deter-mine whether t he next ho p is within communication range.The hello messages may list the o ther nodes from which amobile has heard, thereby yielding greater knowledge of net-work connectivity.Dynamic Source Routing - The Dynamic Source Routing(DSR) protocol presented in [8] is an on-dema nd routingprotocol that is based on t he concept of source routing.Mobile nodes are required to maintain route caches thatcontain the source routes of which the mobile is aware.Entries in the ro ute cache are continually updated as newroutes are learned.The protocol consists of two major phases: route discoveryand route maintenance. When a mobile node has a packet t osend to some destination, it first consults its route cache todetermine whether it already has a route to the destination. Ifit has an unexpired route to the destination, it will use thisroute to send the packet. On the other hand, if the node doesnot have such a r oute , it initiates ro ute discovery by broad-casting a route request packet. This route request contains theaddress of the destinat ion, along with the source nodesaddress and a unique identification number. Each nodereceiving the packet checks whether it knows of a route to thedestination. If it does not, it adds its own address to the routerecord of the packet and then forwards the packet along itsoutgoing links. To limit the number of route requests propa-gated on the outgoing links of a node, a mobile only forwardsthe route request if th e request has not yet been seen by themobile and if the mobiles address does not already appear inthe rou te record.A route reply is generated when the route request reacheseither t he destination itself, or an intermediat e node whichcontains in its route cache an unexpired route to the destina-tion [9]. By the time the packet reaches either the destinationor such an intermediate node, it contains a rou te recordyielding the sequence of hops taken. Figure 4a illustrates theformation of the route record as the route request propagatesthrough the network. If the node generating the route reply

is the destination, it places the route record contained in theroute request into the route reply. If th e responding node isan intermediate node, it will append its cached route to theroute record and then generate the route reply. To return theroute reply, the responding node must have a route to theinitiator. If it has a route to the initiator in its route cache, itmay use that route. Otherwise, if symmetric links are sup-ported, the node may reverse the route in the route record. Ifsymmetric links ar e not supported, the node may initiate itsown route discovery and piggyback the rout e reply on thenew route request. Figure 4b shows the transmission of theroute reply with its associated route record back t o thesource node.Route maintenance is accomplished through the use ofroute error packets and acknowledgments. Route error packetsare genera ted at a node when the data link layer encounters afatal transmission problem. When a route error packet isreceived, the hop in error is removed from the nodes ro utecache and all routes containing the hop are truncated at thatpoint. In addition to route error messages, acknowledgmentsare used t o verify the correct ope ration of th e route links.Such acknowledgments include passive acknowledgments,where a mobile is able to hear the next hop forwarding thepacket along the route.TemporaZly Ordered Routing Algorithm - The TemporallyOrder ed Rout ing Algorithm ( TORA ) is a highly adaptiveloop-free distributed routing algorithm based on the conceptof link reversal [lo] . TORA is proposed to operate in a high-ly dynamic mobile networking environment. It is source-initi-ated and provides multiple routes for any desiredsource/destination pair. The key design concept of TORA isthe localization of cont rol messages to a very small set ofnodes nea r the occurrence of a topological change. Toaccomplish this, nodes need to maintain routing informationabout adjacent (one-hop) nodes. The protocol performs threebasic functions:Route creation- Route maintenanceRoute erasure

During the route creati on and maintenance phases,nodes use a height metric to establish a directed acyclicgraph (DAG) rooted at th e destination. Thereafter, linksare assigned a direction (upst ream or downstream) basedon th e relative height metric of neighboring nodes, asshown in Fig. Sa. This process of establishing a DA G issimilar to the query/reply process proposed in LightweightMobile Routing (LMR) [l l] . In times of node mobility theDAG route is broken, and route maintenance is necessaryto reestablish a DAG rooted at the same destination. Asshown in Fig. 5b, upon fai lure of th e last downstream link,



5/10

a node generates a new reference level which results in thepropagation of that reference level by neighboring nodes,effectively coordinating a structured reaction to the failure.Links are reversed to reflect the change in adapting to thenew reference level. This has t he same effect as reversingthe dir ection of one or mo re links when a node has nodownstream links.Timing is an important factor for TORA because theheight metric is dependent on the logical time of a link fail-ure; TO RA assumes that all nodes have synchronized clocks(accomplished via an external time source such as the GlobalPositioning System). TORAs metric is a quintuple comprisingfive elements. namely:Logical time of a link failureThe unique ID of the node that defined the new reference

A reflection indicator bitA propagation ordering parameterThe unique ID of the nodeThe first th ree elements collectively represent th e refer-ence level. A new reference level is defined each time anode loses ilrs last downstream link due to a link failure.

level

r-

I igure 5. a ) Route creation (showing link direction assign-ment); b ) route maintenance (showing the link reversal phe -nomenon) in TOM .

TORAs rou te e rasure phase essentially involves flooding abroadcast clear packe t (CLR) throughout the network toerase invalid routes.In TOR A ther e is a potential for oscillations to occur,especially when multiple sets of coordinating nodes are con-currently detecting partitions, erasing routes, and buildingnew routes based on each other. Because TORA uses inter-nodal coordination, its instability problem is similar to t hecount-to-infinity problem in distance-vector routing pro to-cols, except that such oscillations a re temporary a nd routeconvergence will ultimately occur.Associativity-Based Routing - A totally different approachin mobile routing is proposed in [12]. Th e Associativity-Based Routing (ABR) protocol is free from loops, deadlock,and packet duplicates, and defines a new routing metric forad hoc mobile networks. This metric is known as the degreeof association stability. In ABR, a route is selected based onthe degree of association stability of mobile nodes. Eachnode periodically generates a beacon to signify its existence.When received by neighboring nodes, this beacon causestheir associativity tables to be up dated. F or each beaco nreceived, the associativity tick of the current node withrespect to the beaconing node is incremented. Associationstability is defined by connection stability of one node withrespect to another node over time and space. A high degreeof association stability may indicate a low state of no demobility, while a low degree may indicate a high state ofnode mobility. Associativity ticks a re reset when t he neigh-bors of a node o r the no de itself move out of proximity. Afundamental objective of ABR is to derive longer-livedroutes for ad hoc mobile networks.

Route discoveryRoute reconstruction (RRC)Route deletionThe route discovery phase is accomplished by a broad-c a s t q u e r y a n d a w a i t - r e p l y (BQ-REPLY) cycle. A n o d edesiring a route broadcasts a B Q message in search ofmobiles that have a route to the dest ination. All nodesreceiving the query (that a re not the destination) a ppendtheir addresses and their associativity ticks with their neigh-bors along with QoS information to the query packet. Asuccessor node erases its upstream node neighbors associa-tivity tick entries and retains only the entry concerned withitself and its upstream node. In this way, each resultantpacket arriving at th e destination will contain th e associativ-ity ticks of the nodes along the route to the destination. Thedestination is then able to select the best route by examin-ing the associativity ticks along each of the p aths. W henmultiple paths have the same overall de gree of associationstability, the r oute with the minimum n umber of hops isselected. The destination then sends a REPL Y packet backto t he source a long th is pa th . Nodes p ropagat ing theREPLY mark their routes as valid. All other routes remaininactive, and the possibility of duplicat e packets arriving atthe destination is avoided.RRC may consist of partial route discovery, invalid routeerasure, valid route updates, and new route discovery, depend-ing on which node(s) along the route move. Movement by thesource results in a new BQ-REPLY process, as shown in Fig.6a. The RN[ l] message is a route notification used to eras ethe rout e entries associated with downstream nodes. Whenthe destination moves, the immediate upstream node erasesits route and determines if the node is still reachable by alocalized query (LQ[HI) process, where H refers to the hopcount from the upst ream node to the destination (Fig. 6b). If

The three phases of ABR are:



6/10

the destination receives the LQ packet, itREPLYs with the best partial route; other-wise, the initiating node t imes out and t heprocess backtracks to th e next upstreamnode. Here an RN[O] message is sent to thenext upstream node to erase the invalidroute and inform this node that it shouldinvoke the LQ[H] process. If this processresults in backtracking more than halfway tothe source, the LQ process is discontinuedand a new BQ process is initiat ed at thesource.When a discovered rout e is no longerdesired, the source node initiates a routedelete (RD ) broadcast so that all nodes along the rout eupdate their routing tables. The RD message is propagated bya full broadcast, as opposed to a directed broadcast, becausethe source node may not be aware of any route node changesthat occurred during RRCs.Signal Stabil i ty Routing - Another on-demand protocol isthe Signal Stability-Based Adaptive Routing protocol (SSR)presented in [13]. Unlike the algorithms described so far, SSRselects routes based on the signal strength between nodes anda nodes location stability. This route selection criteria has theeffect of choosing routes tha t have st ronger connectivity.SSR can be divided into two cooperative protocols: theDynamic Routing Protocol (DR P) and the Static RoutingProtocol (SRP).

W Figure 6.Route maintenance or source and destination movement in ABR.The D RP is responsible for the maintenance of the Sig-nal Stability Table (SST) and Routing Table (RT). Th eSST records the signal strength of neighboring nodes,which is obtained by periodic beacons from t he link layerof each neighboring node. Signal streng th may be recordedas either a strong or weak channel. All transmissions arereceived by, and processed in, the DRP. After updating allappropriate table entries, the DRP passes a received packetto the SRP.The SRP processes packets by passing the packet up thestack if it is the intended receiver or looking up the destina-tion in the RT and then forwarding the packet if it is not. Ifno entry is found in the RT for the destination, a route-searchprocess is initiated to find a route. Route requests are propa-gated throughout the network, but are only forwarded to thenext hoD if thev are received over

Comm unication com piexity (link additiordfailure) O h = N) = PI)

strong ihanne ls and have not beenprevioudy proccsscd (to prcvcntlooping). Thc dcstination chootcs

O(x = AI)I Time complexity (link addition/failure) I

Loop-free YeS Yes Yes, but notinrtantanews

I Routing philosophy I Flat IHierarchical ]Flat * I

Number of required tables TWO Two FourI Multicast capability I No ]NO** INo I

! Updates transmitted t oI

stability, since the packetsarcM g h h N e i g h b o r s ~ ~ * h b o ~ dropped at a node i f they havcarrived over a weak channel. I fluster head

1 Frequency of update transmissions IPeriodically APe riodic ally IPeriodically Iand as needednd as need; Utilizes he llo messages Yes No

thefirstarriving route-search pack-et t o send back because it is mostprobable that t he packet arrivedover the shortest and/or least con-gested path. The DRP then revers-es the selected route and sends aroute-reply message back to the ini-tiator. The DRP of the nodes alongthe path update their RTs accord-ingly.Route-search packets arriving atthe destination have necessarilychosen the path of strongest signal

YeS

* While WRP uses fla t addressing, it can be used hierarchically [15].** The pro toco l itself currently does not support m ulticast; however, there is a separateprotocol described in [16], which runs on top of CGSR and provides multicast capability.

I Utilizes seauence numbers IYes I es lyes I

another route-search process tofind a new path to the destination.The Source sends an mes-sage to notify all nodes of the bro-

1 Critical nodes INo l c i s t er headdNo IAbbreviations:N = Number of nodes in the networkd = Network diameterh = Height of routing treex = Number of nodes affected by a topological change

ther e is no route-reply messagereceived at the source within a spe-cific timeout period, the s ourcechanges the PR EF field in theheader t o indicate that weak chan-nels are acceptable, since these maybe t he only links over which th epacket can be propagated.When a failed link is detectedwithin the network, the intermedi-ate nodes send an err or message tothe source indicating which channelhas failed. The source then initiates



7/10

ComparisonsThe following sections provide comparisons of the previouslydescribed routing algorithms. The next section comparestable-driven protocols, and another section compares on -demand protocols. A l ater section presents a discussion ofthe two classes of algorithms. In Tables 1 and 2 , time com-plexity is defined as the number of steps needed to perform aprotocol operation, and communication complexity is thenumber of messages needed to perform a protocol operation[ l l ,141. Also, the values for these metrics represent worst-case behavior.

Comm unication complexity (postfailure) O ( 2 N ) O(zN) O ( W I

Table-Driven Protocols

Loop-free

The discussion here is based on Table 1. As stated earlier,DSDV routing is essentially a modification of the basic Bell-man-Ford routing algorithm. The modifications include theguarantee of loop-free routes and a simple route updateprotocol. While only providing on e path to any given desti-nation, DSDV selects the shortest path based on the num-ber of hops t o the destination. DSDV provides two types ofupdate messages, one of which is significantly smaller thanthe other . T he smaller update message can be used forincremental updates so that t he entir e routing table neednot be transmitted for every change in network topology.However, DSDV is inefficient because of the requ irementof periodic update transmissions, regardless of the numberof changes in the network topology. This effectively limitsthe number o f nodes that can connect to the network sincethe overhead grows as O(n2 ) . n CGSR, DSDV is used as

Yes YeS Yes Yes Yes

the underlying routing protocol. Routing in CGSR occursover cluster heads and gateways. A cluster head table is nec-essary in addition to th e routing table. O ne advantage ofCGSR is that several heuristic methods can be employed toimprove the protocols performance. T hese me thods includepriority token scheduling, gateway code scheduling, andpath reservation [4].The WR P protocol differs from the othe r protocols inseveral ways. W RP requires each nod e to maintain fourrouting tables. This can lead to substantial memory require-ments, especially when the number of nodes in the networkis large. Furthermo re, the WR P protocol requires t he useof hello packets whenever there are no recent packet trans-missions from a given node. Th e hello packets consumebandwidth an d disallow a node t o ent er sleep mode. How-ever, although it belongs to t he class of path-finding algo-rithms, WRP has an advantage over othe r path-findingalgorithms because it avoids the problem of creating tem-porary routing loops that these algorithms have through theverification of predecessor information, as described in anearlier section.Having discussed the opera tion and characteristics ofeach of the existing table-driven routing protocols, it isimportant to highlight the differences. During link failures,WRP has lower time complexity than DSDV since it onlyinforms neighboring nodes about link status changes. Dur-ing link additions, hello messages are used as a presenceindicator such that t he routing table entry can be updated.Again, this only affects neighboring nodes. I n CGS R,because routing performance is dependent o n the s tatus of

Beaconing requirements No No No Yes Yes

1 Comm unication complexity (initialization) IO(2N) I O(2N) I O(2N) IOW +Y ) IO(N + Y ) I

Routes maintained in Ioute table Route cache Route table Route table Route table

I Routing philosophy IFlat I Flat I Flat I Flat IFlat I

Route reconfiguration methodology Erase route ; Erase rou te; Link reversal; Localized Erase route;notify source notify source route repair broadcast query notify source

I Multicast capab ility IYes INo INO** I No IN0 II Multip le rout e possibilities INo IYes I Yes INo IN 0 I1 Utilizes route cach dtable expiration timers lYes I No INo INo INo I

Routing metric Freshest and Shortest pa th Shortest pa th Associativity and Associativity andshortest path1 1 1 1 others*** 1 1hortest path and stability

W Table 2. Comparisons of the characteristics of source-initiated on-dema nd ad hoc routingprotocols.



8/10

specific nodes (cluster head, gateway, o r normal nodes),time complexity of a link failure associated with a clusterhead is higher than DSDV, given the additional time need-ed to perform cluster head resel ection. Similarly, thisapplies to the case of link additions associated with the clus-ter head. Ther e is no gateway selection in CGSR since eachnode de clares it is a gateway node t o its neighbors if it isresponding to multiple r adio codes. If a gateway nodemoves out of range, the routing protocol is responsible forrouting the packet to another gateway.In terms of communication complexity, since DSDV,CGSR, and WRP use distance vector shortest-path routing asthe underlying routing protocol, they all have the same degreeof complexity during link failures and additions.Sou rce-In itia ed On-Deman d Routing Protocols

Table 2 presents a comparison of AODV, DSR, TORA,ABR, and SSR.AODV employs a route discovery procedure similar toDSR; however, there are a couple of important distinctions.The most notable of these is that the overhead of DSR ispotentially larger than that of AODV since each DSR packetmust carry full routing information, whereas in AODV pack-ets need only contain the destination address. Similarly, theroute replies in DSR a re larger because they contain theaddress of every node along the route, whereas in AOD Vroute replies need only carry the destination IP address andsequence number. Also, the memory overhead may be slight-ly grea ter in DSR because of the need t o remember fullroutes, as opposed to only next hop information in AODV. Afurther advantage of AODV is its support for multicast [17].None of th e othe r algorithms considered in this article cur-rently incorporate multicast communication. On the down-side, AODV requires symmetric links between nodes, andhence cannot utilize routes with asymmetric links. In thisaspect DSR is superior, since it does not require the use ofsuch links and can utilize asymmetric links when symmetriclinks are not available.The DSR algorithm is intende d for networks in whichthe mobiles move at moderate speed with respect to packettransmission latency [8]. Assumptions the algorithm makesfor operation are that the network diameter is relativelysmall and that the mobile nodes can enable a promiscuousreceive mode, whereby every received packet is delivered tothe network driver software without filtering by destinationaddress. An advantage of DSR over some of the other on-demand protocols is that DSR does not make use of period-ic routing advertisements, thereby saving bandwidth andreducing power consumption. Hence, the protocol does notincur any overhead when there are no changes in networktopology. Additionally, DSR allows nodes to keep multipleroutes to a destination in their cache. Hence, when a linkon a rout e is broken, the source node can check its cachefor anoth er valid route. If such a route is found, routereconstruction does not need to be reinvoked. In this case,rout e recovery is faster than in many of the ot her on-demand protocols. However, if t here a re no additionalroutes to t he destination in the source nodes cache, routediscovery must be reinitiated, as in AODV, if the route isstill required. On the ot her hand, because of the smalldiameter assumption and the source routing requirement,DSR is not scalable to large networks. Furthermore, as pre-viously stated, the need to place t he entire rout e in bothroute replies and data packets causes greater control over-head than in AODV.TO RA is a link reversal algorithm tha t is best suitedfor networks with large dense populations of nodes [lo] . Part

of the novelty of T ORA stems from its creation of DAGS toaid route establishment. One of the advantages of TORA isits support for multiple routes. TORA and DSR ar e the onlyon-demand protocols considered here which retain multipleroute possibilities for a single source/destination pair. Routereconstruction is not necessary until all known routes to adestinat ion are deemed invalid, and hence bandwidth canpotentially be conserved because of the necessity for fewerroute rebuildings. Anothe r advantage of TO RA is its sup-port for multicast. Although, unlike AODV, TORA does notincorporate multicast into its basic operation, it functions asthe underlying protocol for the Lightweight Adaptive Multi-cast Algorithm (LAM), and together t he two protocols pro-vide multicast capability [18]. TORAs reliance onsynchronized clocks, while a novel idea, inherently limits it sapplicability. If a node does not have GPS or some otherexternal time source, it cannot use t he algorithm. Addition-ally, if the external time source fails, the algorithm will ceaseto operate. Furthermore, route rebuilding in TOR A may notoccur as quickly as in the other algorithms due to the poten-tial for oscillations during this period. This can lead topotentially lengthy delays while waiting for the new routes tobe determined.ABR is a compromise between broadcast and point-to-point routing, and uses the connection-oriented packet for-warding approach. Route selection is primarily based on theaggregated associativity ticks of nodes along the pa th. Hence,although the resulting path does not necessarily result in thesmallest possible number of hops, the path tends to be longer-lived than oth er routes. A long-lived route requires fewerroute reconstructions and therefore yields higher throughput.Another benefit of ABR is that, like the other protocols, it isguaranteed to be free of packet duplicates. The reason is thatonly the best rou te is marked valid, while all other possibleroutes remain passive. ABR, however, relies on the fact thateach node is beaconing periodically. The beaconing intervalmust be short enough to accurately reflect the spatial, tempo-ral, and connectivity state of the mobile hosts. This beaconingrequirement may result in addit ional power consumption.However, experimental results obtained in [19] reveal that theinclusion of periodic beaconing has a minute influence on theoverall battery power consumption. Unlike DSR, ABR doesnot utilize route caches.The SSR algorithm is a logical descendant of ABR. It uti-lizes a new technique of selecting routes based on the signalstrength and location stability of nodes along the path. As inABR, while the paths selected by this algorithm are not neces-sarily shortest in hop count , they do tend to be more stableand longer-lived, resulting in fewer route reconstructions. Oneof the major drawbacks of the SSR protocol is that, unlike inAODV and DSR, intermediate nodes cannot reply to routerequests sent toward a destination; this results in potentiallylong delays before a rou te can be discovered. Additionally,when a link failure occurs along a path, t he ro ute discoveryalgorithm must be reinvoked from the source to find a newpath to the destination. N o attempt is made to use partialroute recovery (unlike ABR), t hat is, to allow intermediat enodes to attempt to rebuild the route themselves. AODV andDSR also do not specify intermediate node rebuilding. Whilethis may lead to longer route reconstruction times since linkfailures cannot be resolved locally without the intervention ofthe source node, the attempt and failure of an intermedia tenode to rebuild a r oute will cause a longer delay than if thesource node had attempted the rebuilding as soon as the bro-ken link was noticed. Thus, it remains to be seen whetherintermediate node route rebuilding is more optimal thansource node route rebuilding.

IEEE PersonalCommunications April 1999 53


9/10

Available when needed Always ava ilable regardlessofIneedAvailability of routinginformationRouting philosophy Flat Mostly flat, except for CGSR

1 Periodic route updates I Not required IRequired I1Coping with m obility Use localized r out e discoveryas in ABR and SSR

IInform other nodes to achievea consistent routing table i

Signaling traffic generated Grows with increasing mobility Greater than that of on-of active rou tes (as in ABR) I emand routing IQuality of service support---___.-.___

Few can support QoS, although Mainly shortest path as themost support shortestpath 00s metric Ii-__..___ .___ ..-Table 3. Overall comparisons of on-demand versus table-driven routingprotocols.Table-Driven vs. On -Dem and Routing

As discussed earlier, the table-driven ad hoc routing approachis similar to 1he connectionless approach of forwarding pack-ets, with no regard to when and how frequently such routesare desired. It relies on an underlying routing table updat emechanism that involves the constant propagation of routinginformation. This is not the case, however, for on-demandrouting protocols. When a node using an on-demand protocoldesires a route to a new destination, it will have to wait untilsuch a route can be discovered. On the other hand, becauserouting information is constantly propagated and maintainedin table-driven routing protocols, a route to every other nodein the ad hoc network is always available, regardless ofwhether or not it is needed. This feature, although useful fordatagram traffic, incurs substantial signaling traffic and powerconsumption. Since both bandwidth and battery power arescarce resources in mobile computers, this becomes a seriouslimitation. Table 3 lists some of the basic differences betweenthe two classes of algorithms.Anothe r considerat ion is whether a flat or hierarchicaladdressing scheme should be used. All of the protocols con-sidered here, except for CGSR, use a flat addressing scheme.In [20] a discussion of the two addressing schemes is present-ed. While flat addressing may be less complicated and easierto use, there are doubts as to its scalability.

Ap p ica tions and Cha Zleng esAkin to packet radio networks, ad hoc wireless networks havean important role to play in military applications. Soldiersequipped with multi mode mobile communicators can nowcommunicate in an ad hoc manner without the need for fixedwireless base stations. In addition, small vehicular devicesequipped with audio sensors and cameras can be deployed attargeted regions to collect important location and environ-mental information which will be communicated back to aprocessing node via ad hoc mobile communications. Ship-to-ship ad hoc mobile communication is also desirable since itprovides alter nate communication paths without reliance onground- or space-based communication infrastructures.Commercial scenarios for ad hoc wireless networks include:Conferences/meetings/lecturesEmergency servicesLaw enforcementPeople today attend meetings and conferences with theirlaptops, palmtops, and notebooks. It is therefore attractive tohave instant network formation, in addition to file and infor-mation sharing without the presence of fixed base stations andsystems administrators. A presenter can multicast slides andaudio to intended recipients. Attendees can ask questions and

interact on a commonly sharedwhiteboard . Ad hoc mobile com-munication is particularly useful inrelaying information (status, situa-tion awareness, etc.) via data, video,and/or voice from one rescue teammember to ano ther over a smallhandheld or wearable wirelessdevice. Again, thi s applies t o lawenforcement personnel as well.Current challenges for ad hocwireless networks include:*Multicast [21]*QoSsupportPower-aware routing [22]Location-aided routing [23]As mentioned above, multicastis desirable to support multiparty wireless communications.Since the multicast tree is no longer sta tic (i.e., its topologyis subject t o change over time), th e multicast routing proto-col must be able to cope with mobility, including multicastmembership dynamics (e.g., leave and join). In terms ofQoS, it is inadequate to consider QoS merely at the networklevel without considering the underlying media access con-trol layer [24]. Again, given the problems associated with thedynamics of nodes, hidden terminal s, and fluctuating linkcharacteristics, suppor ting end-to-end Q oS is a nontrivialissue that requires in-depth investigation. Currently, there isa trend toward a n adaptive QoS approach instead of theplain resource reservation method with hard QoS guaran-tees. Another important factor is the limited power supply inhandheld devices, which can seriously prohibit packet for-warding in an ad hoc mobile environment. Hence, routingtraffic based on nodes power metrics is one way to distin-guish routes that are more long-lived than ot hers. Finally,instead of using beaconing or broadcast search, location-aided routing uses positioning information to define associ-ated regions so t ha t the routing is spatially or i en ted an dlimited. This is analogous t o associativity-oriented andrestricted broadcast in ABR.Current ad hoc routing approaches have introduced severalnew paradigms, such as exploiting user demand, and the useof location, power, and association parameters. Adaptivity andself-configuration are key features of these approaches. How-ever, flexibility is also important. A flexible ad hoc routingprotocol could responsively invoke table-driven andlor on-demand approaches based on situations and communicationrequirements. Th e toggle between these two approac hesmay not be trivial since concerned nodes must be in syncwith the toggling. Coexistence of both approaches may alsoexist in spatially clustered ad hoc groups, with intraclusteremploying the table-driven approach and intercluster employ-ing the demand-driven approach or vice versa. Further work isnecessary to investigate the feasibility and performance ofhybrid ad hoc routing approaches. Lastly, in addition t o theabove, further research in the areas of media access control,security, service discovery, and Internet protocol operability isrequired before the potential of ad hoc mobile networking canbe realized.

Concl usionIn this article we provide descriptions of several routingschemes proposed for ad hoc mobile networks. We alsoprovide a classification of the se schemes according to t herouting strategy (i.e., table-driven a nd on-demand). Wehave presented a comparison of these two categories of



10/10

routing protocols, highlighting their features , differences,and characteristics. Finally, we have identified possibleapplications and challenges facing ad hoc mobile wirelessnetworks. While it is not clear that any particular algorithmor class of algorithm is the best for all scenarios, each pro-tocol has definite advantages and disadvantages, and is wellsuited for certai n situations. The field of ad hoc mobilenetworks is rapidly growing and changing, and while t hereare still many challenges that need to be met, it is likelythat such networks will see widespread use within th e nextfew years.Acknawledgm n s

We would like to thank C.-C. Chiang, J . J. Garcia-Luna-Aceves, David Maltz, Vincent Park, and Charles Perkins fortheir assistance in ensuring accurate descriptions of the opera-tion of their protocols.References

[I] J. Jubin and J. Tornow, "The DARPA Packet Radio Network Protocols,"Proc. /E, vol. 75,no. 1, 1987,pp, 21-32.[2] . E. Perkins and P. Bhagwat, "High ly Dynamic Destination-SequencedDistance-Vector Routing (DSDV) for Mobi le Computers," Comp. Com-mun. Rev., Oct. 1994,pp. 234-44.[3] . R. Ford Jr. and D. R. Fulkerson, Flows in Networks, Princeton Univ.Press, 1962.[41C.-C. Chiang, "Routing in Clustered Multihop, Mobile Wireless Networks withFading Channel," Proc. /E SlCON '97,Apr. 1997,pp. 197-21 .[5]. Murthy and J. J. Garcia-Luna-Aceves, "An Efficient Routing Protocol forWireless Networks," ACM Mobile Networks and App. 1..Special Issue onRouting in Mobile Communication Networks, Oct. 1996,pp. 183-97.

[6] . S.Tanenbaum, Computer Networks, 3rd ed., Ch. 5, Englewood Cliffs,NJ: Prentice Hall, 1996,pp. 357-58.[7] . E. Perkins and E. M. Royer, "Ad-hoc On-Demand Distance Vector Routing,"Proc. 2nd /E Wksp. Mobile Comp. Sys. and Apps., Feb. 1999,pp. 9C100.[8] . B . Johnson and D. A. Maltz, "Dynamic Source Routing in Ad-HocWireless Networks," Mobi le Computing, T. lmielinski and H. Korth, Eds.,Kluwer, 1996, p. 153-81.[9] . Broch. D. B. Johnson, and D. A. Maltz, "The Dynamic Source RoutingProtocol for Mobile Ad Hoc Networks," IETF Internet draft, draft-ietf-manet-dsr-01 txt, Dec. 1998 work in progress).[lo] . D. Park and M. S. Corson, "A Highly Adaptive Distributed RoutingAlgorithm for Mobile Wireless Networks," Proc. lNFOCOM'97,Apr. 1997.[ l l ] M. S. Corson and A. Ephremides, "A D istributed Routing Algori thm for

Mobile Wireless Networks," ACMIBaltzer Wireless Networks 1.. vol. 1,no. 1, Feb. 1995,pp. 61-81.[12] -K. Toh, "A Novel Distributed Routing Protocol To Support Ad-HocMobile Computing," Proc. 7996 E 75th Annual lnt'l. Phoenix Conf.Comp. and Commun., Mar. 1996,pp. 480-86.[13] . Dube et al., "Signal Stability based Adaptive Routing (SA ) for Ad-Hoc Mobi le Networks," / PerS. Commun., Feb. 1997, p. 36-45.[14] -K. Toh, "Associativity-Based Routing for Ad-Hoc Mobile Networks,"Wireless Pers. Commun., vol. 4,no. 2,Mar. 1997,pp. 1-36.

[151 . Murthy and J. J. Garcia-Luna-Aceves, "Loop-Free Interne t RoutingUsing Hierarchical Routing Trees," Proc. lNFOCOM '97, Apr. 7-1 1, 1997.[161C.-C. Chiang, M. Gerla, and L. Zhang, "Adaptive Shared Tree Multicast nMobile Wireless Networks," Pnx. GLOBECOM'98. Nov. 1998.pp. 1817-22.[171C. E. Perkins and E. M. Royer, "Ad Hoc On Demand Distance Vector(AODV) Routing," IETF Internet draft , draft-ietf-manet-aodv-O2,txt, Nov.1998 work in progress).[I81 . Jiand M. S.Corson, "A Lightweight Adaptive Multicast Algorithm,"Proc. GLOBECOM '98. Nov. 1998.pp. 1036-42.[191C.-K. Toh and George Lin, "Implementing Associativity-Based Routingfor Ad Hoc Mobile Wireless Networks," Unpublished article, Mar. 1998.1201 D. Baker et al., "Flat vs. Hierarchical Network Control Architecture,"ARO/DARPA Wksp. M obile Ad-Hoc Networking; http://www.isr.umd.edu/CourseWorkshops/MANET/program.html,Mar. 1997.[211M. Gerla, C.-C. Chiang, and L. Zhang, "Tree Multicast Strategies inMobile, Multi hop Wireless Networks," ACMIBaltzer Mobi le Networksand Apps. 1..1998.1221 S.Singh, M. Woo, and C. S. Raghavendra, "Power-Aware Routing inMobile Ad Hoc Networks," Proc. ACMll MOBlCOM'98,Oct. 1998.[231 . B. KO and N. H. Vaidya, "Location-Aided Routing (IAR) in Mobile AdHoc Networks," Proc. ACMllE MOBlCOM '98,Oct. 1998.[241C. R. Lin and M. Gerla, "MACNPR: An Asynchronous Multimedia Multi-hop Wireless Network," Proc. / E lNFOCOM'97,Mar. 1997.

BiographiesELIZABAETH M. ROVER ([email protected]) received her M.S. degreein electrical and computer engineering from the University of California,Santa Barbara in December 1997,and her B.S. degrees in computer scienceand applied mathematics from Florida State University in April 1996.She iscurrently a Ph.D. student in the Electrical and Computer EngineeringDepartment at the University of California, Santa Barbara. At UCSB she is amember of the Computer Networks and Distributed Systems Laboratory.Her research interests include routing and multicast for ad hoc mobile net-works. She is the recipient of a National Science Foundation Graduate Fel-lowship and a University of California Doctoral Scholar Fellowship.CHAI-KEONGTOH [MI ([email protected]) received his Diploma in electronicsand communication engineering with a Certificate of Merit award from theSingapore Polytechnic n 1986, is B.Eng. degree in electronics engineeringwith first class honors from the University of Manchester Institu te of Scienceand Technology (UMIST) in 1991,and his Ph.D. degree in computer sciencefrom the Computer Laboratory, University of Cambridge, England in 1996.Hefounded and chaired the Mobile Special Interest Group (Mobile SIC) from1994 o 1996.Before oining Cambridge University, he was a network special-ist, R&D engineer and technical staff member. At Cambridge, he was an Hon-orary Cambridge Commonwealth Trust Scholar and a King's CollegeCambridge Research Scholar. He authored the book Wire/ess ATM and Ad HocNetworks: Protocols and Architectures (Kluwer, 1996).He serves on the edito-rial board for ACM Mobile Computing and Communications Review, /E Net-work, and Personal Technologies Journal (Springer Verlag). He is a member ofI EE , USENIX, ACM, Sigma Xi Honor Society, New York Academy of Science,American Association for the Advancement of Science(a),nd is a Fellowof the Cambridge Philosophical Society and Cambridge Commonwealth Soci-ety. He is currently an assistant professor wit h the School of Electrical andComputer Engineering a t the Georgia Institute of Technology, Atlanta, direct-ing the Mobile Mult imedia and High Speed Networking Laboratory.

IEEE PersonalCommunications April 1999 55

a review on current routing protocols in wireless ad hoc netwrks

Documents