IACSIT International Journal of Engineering and Technology, Vol.2, No.4, August 2010
ISSN: 1793-8236
394
Abstract—Mobile ad hoc networks (MANETs) consist of a
collection of wireless mobile nodes which dynamically
exchange data among themselves without the need of fixed
infrastructure or a wired backbone network. Due to limited
transmission range of wireless network nodes, multiple hops
are usually needed for a node to exchange information with
any other node in the network. Thus routing is a crucial issue
in the design of MANET. On-demand routing protocols for
mobile ad hoc networks discover and maintain only the needed
routes to reduce routing overheads. They use a flood-based
discovery mechanism to find routes when required. Since each
route discovery incurs high overhead and latency, the
frequency of route discoveries must be kept low for on-
demand protocols to be effective. The wide availability of
wireless devices requires the routing protocol should be
scalable. But, as the size of the network increases the on-
demand routing protocols produce poor performance due to
large routing overhead generated while repairing route breaks.
The proposed multipath routing scheme provides better
performance and scalability by computing multiple routes in a
single route discovery. Also, it reduces the routing overhead by
using secondary paths. This scheme computes combination of
the node-disjoint path and fail-safe paths for multiple routes
and provides all the intermediate nodes of the primary path
with multiple routes to destination.
Index Terms—Mobile ad hoc networks; Multipath routing;
Fail-safe multiple path; Primary path; Secondary path
I. INTRODUCTION
Mobile Ad hoc Networks (MANETs) are autonomous
networks, which operate without any fixed infrastructure or
wired backbone. In MANETs, nodes typically communicate
over multiple hops while the intermediate nodes act as
routers by forwarding data. Because of mobility and limited
battery power of nodes, topology of ad hoc network is
highly dynamic. Hence routing protocols should adapt to
such dynamic nature and continue to maintain connection
between the communicating nodes even if path breaks due
to mobility and or node failures.
The objective of this paper is to develop multiple routes
in order to improve scalability. Byfinding multiple paths in
a single route discovery, reduce the routing overhead
incurred in maintaining the connection between source and
destination nodes. The secondary paths can be used to
transmit data packets, in case the primary path fails due to
node mobility or battery failure, which avoids extra
overhead generated by a fresh route discovery. These
multiple paths are more advantageous in larger networks,
where he number of route breaks are high.
When a source node needs to send data to destination and
does not have a valid path to destination, it starts a timer
and relays a route request (RREQ) for destination with
unique route request identifier. When source node receives
a feasible reply for the destination, it updates its route table
and starts sending a data packet. If the timer expires in
between, then source node increments the route request
identifier and initiates a new request for the destination.
Multipath routing can increase end-to-end throughput and
provide load balancing in MANETs by the use of multiple
paths. The concept of multipath routing motivated to design
a multipath routing for mobile ad hoc networks.
1. To avoid the overhead of additional route discovery
attempts.
2. To minimize the routing overhead by the use of
secondary paths.
3. To reduce the route error transmission during route
break recovery.
II. RELATED WORKS
In this section, we have given a brief review of routing
protocols which is for multipath routing. Mobile ad hoc
networks (MANETs) are characterized by a dynamic
topology, limited channel bandwidth and limited power at
the nodes. Because of these characteristics, paths
connecting source nodes with destinations may be very
unstable and go down at any time, making communication
over ad hoc networks difficult. On the other hand, since all
nodes in an ad hoc network can be connected dynamically
in an arbitrary manner, it is usually possible to establish
more than one path between a source and a destination.
When this property of ad hoc networks is used in the
routing process, then it is called multipath routing.
In most cases, the ability of creating multiple routes from
a source to a destination is used to provide a backup route.
When the primary route fails to deliver the packets in some
way, the backup is used. This provides a better fault
tolerance and efficient recovery from route failures.
Multiple paths can also provide load balancing and route
failure protection by distributing traffic among a set of paths.
Multiple paths between a source and a destination can be
disjoint in two ways: (a) link-disjoint paths and (b) node-
disjoint paths. Node-disjoint paths do not have any nodes in
common, except the source and destination hence they do
not have any links in common .Link-disjoint paths, in
contrast, do not have any links in common.
Many on-demand multipath routing protocols have been
proposed for mobile ad hoc networks, including Split
Multipath Routing (SMR), Multipath Dynamic Source
Routing (Multipath DSR), Temporally Ordered Routing
Algorithm (TORA), Routing On-demand Acyclic Multipath
(ROAM), Ad hoc On-demand Multipath Distance Vector
(AOMDV), AODV-BR Ad hoc On-demand Distance
Vector Backup Routing (AODV-BR) and Cooperative
Packet Caching and Shortest Multipath (CHAMP). SMR
and multipath DSR are based on source routing and are
An Extended AODV Protocol for Multipath
Routing in MANETs
N.Jaisankar
1
and R.Saravanan
2
IACSIT International Journal of Engineering and Technology, Vol.2, No.4, August 2010
ISSN: 1793-8236
395
based on DSR while TORA, ROAM, AOMDV are
distance-vector based. AODV-BR and AOMDV routing
protocols are based on AODV.
Sung-Ju Lee and Mario Gerla proposed AODV-BR [2]
routing protocol. The AODV-BR protocol uses the route
discovery process as AODV [1]. When a source needs a
route to a destination, and there is no route to that
destination in its route cache, it searches a route by flooding
a route request (RREQ) packet. Each of these packets has a
unique ID so intermediate nodes can detect and drop
duplicates. When an intermediate node receives a RREQ, it
records the previous hop and the source node information
and then broadcasts the packet or sends a route reply (RREP)
packet back to the source if a route to the desired
destination is known. The destination sends a RREP via the
selected route when it receives the first RREQ or later
RREQs that traversed a better route (with fewer hops).
The alternate route creation part is established during the
RREP phase, and uses the nature of wireless
communications. When a node that is not part of the
selected route overhears a RREP packet not directed to it. It
records the sending neighbor as the next hop to the
destination in its alternate route table. In this way a node
may receive numerous RREPs for the same route, select the
best route among them and insert it into the alternate route
table.
When an RREP finally reaches the source of the route, a
primary route between that source and destination has been
established. All the nodes that have an alternate route to the
destination in their alternate route table form a fish bone.
The properties of AODV-BR are is an extension of AODV.
They floods RREQs with unique ID so duplicates can be
discarded. Each node maintains backup route(s) in an
alternate table. No multiple complete routes available. No
multiple route(s) information known at source.
Mahesh K. Marina Samir R. Das proposed AOMDV [3]
routing protocol. Like AODV-BR, the AOMDV uses the
basic AODV route construction process. In this protocol
some extensions are made to create multiple loop-free, link-
disjoint paths. The main idea in AOMDV is to compute
multiple paths during route discovery. It consists of two
components: (i) A route update rule to establish and
maintain multiple loop-free paths at each node. (ii) A
distributed protocol to find link-disjoint paths. In AODV,
when a source needs a route to a destination, it initiates a
route discovery process by flooding a RREQ for destination
throughout the network. RREQs should be uniquely
identified by a sequence number so that duplicates can be
recognized and discarded. Upon receiving a non-duplicate
RREQ, an intermediate node records previous hop and
checks whether there is a valid and fresh route entry to the
destination in routing table. If such case, the node sends
back a RREP to the source if not rebroadcasts the RREQ by
incrementing the hopcount. A node updates its routing
information and propagates the RREP upon receiving
further RREPs only if a RREP contains either a larger
destination sequence number (fresher) or a shorter route
found.
In AOMDV each RREQ, respectively RREP arriving at a
node potentially defines an alternate path to the source or
destination. Just accepting all such copies will lead to the
formation of routing loops. In order to eliminate any
possibility of loops the “advertised hopcount” is introduced.
The advertised hopcount of a node i for a destination d
represents the maximum hopcount of the multiple paths for
d available at i. The protocol only accepts alternate routes
with hopcount lower than the advertised hopcount, alternate
routes with higher or the same hopcount are discarded. The
advertised hopcount mechanism establishes multiple loop-
free paths at every node. These paths still need to be disjoint.
In AOMDV duplicate copies of a RREQ are not
immediately discarded. Each packet is examined to see if it
provides a node-disjoint path to the source. For node-
disjoint paths all RREQs need to arrive via different
neighbor of the source. This is verified with the first hop
field in the RREQ packet and the first hop list for the RREQ
packets at the node. At the destination a slightly different
approach is used, the paths determined are link-disjoint or
node-disjoint. In order to do this, the destination replies up
to k copies of the RREQ, regardless of the first hops. The
RREQs only need to arrive via unique neighbors.
S.Lee and Mario Gerla proposed SMR [4] protocol. It
provides way of determining maximally disjoint paths.
Paths are maximally disjoint when they are node disjoint,
but when there are no node-disjoint paths available, the
protocol minimizes the number of common nodes. Multiple
routes are discovered on demand, one of which is the path
with the shortest delay. The routes established by the
protocol are not necessarily equal in length. Saleem et. al
[10] proposed the model of self-optimized multipath routing
algorithm. Fujian Qin [11] a multipath source routing
protocol with bandwidth and reliability guarantee is
proposed. In the routing discovery pahse, the protocol
selects several multiple alternate paths which meet the QoS
requirements and the ideal number of multipath routing is
achieved to compromise between load balancing and
network overhead. In the routing maintenance phase, it can
effectively deal with route failures similar to DSR.
Furthermore, the per-packet granularity is adopted in traffic
allocation phase. Simulation results show that the proposed
protocol remarkably increases the packet delivery rate and
life-span of network with lower routing overhead. Yuwang
Yang, et.al[12] presents network coding based reliable
disjoint and braided multipath routing (NC-RMR) for
sensor networks, which forms multipath by hop-by-hop
method and only maintains local path information of each
node without establishing end-to-end paths.
III. P
ROPOSED MULTIPATH ROUTING SCHEME
This paper proposes a multipath routing scheme called
Multipath On-demand Routing (MORT), in order to
minimize the route break recovery overhead. This scheme
provides multiple routes on the intermediate nodes on the
primary path to destination along with source node. The
primary path is the first path received by the source node
after initiating the route discovery, which is usually the
shortest path. Having multiple routes at the intermediate
nodes of the primary path, avoid overhead of additional
route discovery attempts, and reduce the route error
transmitted during route break recovery.
IACSIT International Journal of Engineering and Technology, Vol.2, No.4, August 2010
ISSN: 1793-8236
396
Multipath routing protocols work on the principle that
higher performance can be achieved by recording more than
one feasible path. When multiple routes are known, even if
the primary path fails data forwarding can continue
uninterrupted on the alternate available paths without
waiting for a new route to be discovered. In this scheme, the
single-path AODV has been extended for multipath routing.
This scheme is used for infrastructureless networks in which
communication failure occurs frequently and designed to
calculate node-disjoint paths and fail-safe paths. In node –
disjoint path do not have any particular nodes in common,
except the source and destination, whereas fail-safe is a path
between source and destination if it bypasses at least one
intermediate node on the primary path, which is the shortest
path between the source and destination pair. Thus fail-safe
path is different from node-disjoint and link-disjoint paths,
in the sense that fail-safe path can have both nodes and links
in common.
On-demand routing scheme that computes fail safe
multiple paths reduces the route recovery time and path
maintenance overhead more effectively than the node-
disjoint multipath routing scheme. When node-disjoint
multiple paths are used, only the source can correct the
route disconnections, as alternate paths exist only at that
node. In effect, route error packets have to be sent to the
source node for every link break. In large networks, these
error packets are likely to take considerable amount of time
to reach the source node from the point of route break.
Besides, the number of route errors communicated may also
be high, as more number of nodes transmits these packets.
Alternatively, usage of fail-safe paths has the advantage that
route disconnection gets corrected at an intermediate node
itself, thereby reducing the route recovery time and the
number of route error transmissions.
The proposed scheme provides multiple alternative paths
using the combination of the node-disjoint path and fail-safe
paths. This scheme has more alternative paths than node-
joint or link-disjoint paths. Each MANETs node keeps and
maintains tables—routing table, and neighbor node table.
The propsoed scheme has two basic phases:
¾ route discovery
¾ route maintenance
A. Route discovery process
First, to find routes for a destination node, a source node
broadcasts an RREQ packet. When an intermediate node
receives the first RREQ packet, it records a node address in
route request table to relay RREP. When an intermediate
node receives another RREQ packet again, then the node
checks a node list field in the packet. The packet would be
discarded immediately when the field contains the same
node’s IP address that of in the RREQ packet, else stores a
node details into the request received table. After storing the
node details it checks for route to a destination is exist in its
routing table. If this check is passed then creates a RREP
and send to the source using request received table entry. If
not it re-broadcasts the RREQ packet by incrementing the
hopcount. When a RREQ is received by destination node
itself it stores the node address which relayed the RREQ in
the request received table and creates a RREP, updates its
routing table and send the RREP to its upstream nodes using
request received table information.If a node receives a
RREQ for the first time, it searches for a reverse route to the
source. If no reverse route exists, then it will create a new
route.
The extension of the RREP packet structure is given
below:
TABLE 1. RREP PACKET STRUCTURE
The structure of the routing table is given below:
TABLE 2: ROUTE TABLE ENTRY STRUCTURE
The following details of different packet information
have been given below:
The RREQ packet is created based on the IETF format
specification. The fields are:
PacketType: To identify the type of packet
SrcAddr : The node address which generates RREQ
SrcSeqNo : Sequence number of source node
BcastId : Request Id of RREQ
DestAddr : Destination node address
Dest SeqNo : Sequence number of destination node
Hopcount : Number of hops from source
The RREP packet is created with three additional fields.
The format of RREP is:
PacketType : To identify the type of packet
SrcAddr : The node address which generates RREQ
DestAddr : Destination node address
Repgen : Address of the node which generates the
RREP
Mulreply : Is a Boolean value. Set TRUE for the first
Reply
Nodelist : List of node address which relayed the
RREQ
Hopcount : Number of hops to reach the source node
The fields of RERR packet is:
PacketType : To identify the type of packet
IACSIT International Journal of Engineering and Technology, Vol.2, No.4, August 2010
ISSN: 1793-8236
397
NodeAddr : Address of node where link failure is
occurred.
The route table consists of the following information:
DestAddr : Address of the destination node.
RouteList : This filed holds multiple routes with
the
values of nexthop, hopcount, lifetime,
and fullpath.
PrecurList : Holds list of nodes that relayed a RREQ
Packet
In this scheme, the destination is responsible for
discovering primary path, node-disjoint paths and fail-safe
paths from all the received routes as well as defining the
route labels. The destination receives the RREQ for the first
time, which stores the route path of RREQ and sets it with
route label. Then the destination node creates route reply
(RREP) in which route path is included. Once created,
RREP will be unicast to the next hop according to route
path towards the source S and the hop_cnt also incremented
by one at each hop. Hence the intermediate nodes can
forward this packet using path information in RREP. As the
RREP reaches the source the hop count represents the
distance, in hops, of the destination from the source. When
the destination receives a duplicate RREQ, it will compare
route path of RREQ to that of the routing table, then the
path will be selected. The number of multiple paths between
source node S and destination D can be discovered using
selective RREQ forwarding scheme during route discovery
process.
The number of RREP packets generation is limited to
MAX_REPLY. The intermediate node that receives the first
RREP packet forwards it to any neighbors using request
received table that forms a reverse routes toward a source
node and updates its routing table. Routing loop can be
easily avoided by using the node list attached. If the node
receives a delayed RREP packet, it updates routing table
similar to the RREQ extension case, discarding the RREP
packet. In this route accumulation process, nodes are adding
their neighbor node route information as well as which type
of paths are used in the route discovery process. If the
destination nodes don’t have the reverse route, it finds one
new reverse route to the source.
Fig .1 Discovering multiple paths during route discovery
Finally, the fastest RREP for the source node provides a
primary route. The others are examined in the source node
as well as in intermediate nodes, and some of the routes are
accepted as backup routes according to the full path
information. Data transfer begins just after the primary
route is established. When the destination receives the
duplicate RREQ packet, it will compare route path of
RREQ of that routing table. If the source and destination
nodes are same, then the path is said to be a node- disjoint
path and the destination determines it as path type two.
If at least one of intermediate nodes in the route path in
the routing table is different from nodes in the route path of
the RREQ, a route is said to be a fail-safe path and
destination determines it as path type three. After setting
appropriate route label in RREP, the destination sends it to
the source along the path information in it.
As shown in Figure 1, number of multiple paths between
source node S and destination D can be discovered using
selective RREQ forwarding scheme during route discovery
process. After completion of route discovery process, there
will be a primary path <S–N1–N2–N3–D>; two node-
disjoint paths <S–N5–N6–N7–D>, and <S–N10–N11–N12–
D>; and a number of fail-safe paths <S–N5–N2–N7–D>,
<S–N1–N6–N3–D>, <S–N1–N2–N12–D>, <S–N10–N11–
N3–D>.
B. Route maintenance
When a node cannot receive HELLO messages from
neighbors, the node detects link break. If neighbor nodes do
not have any backup routes, the nodes invalidate their
current routing tables and find precursor lists to send RERR
packets to its neighbor nodes. Otherwise, the nodes
immediately change a current route to a backup route.
Avoidance of re-route discovery contributes to reduction of
packet delay and the amount of routing packets in network.
In addition, HELLO packets detecting link failure can
update the backup route expiration timer and extend its life
cycle.
Results and Discussion
The proposed scheme has been implemented in NS2. The
simulation environment consists of different number of
nodes in a rectangular region of varying size. The nodes are
randomly placed in the region and each of them has a radio
range of 150 meters. Five sessions of Constant Bit Rate
flows are developed for data transmission. The random
waypoint model is chosen as the node mobility model.
Simulation time is 300 seconds. Each scenario is simulated
five times and an average is taken for the performance
analysis. The random waypoint model is chosen as the node
mobility model. All data packets are 512 bytes. Table 1
shows the simulation parameters used.
Simulation time 300 seconds
Number of nodes 100 to 1000
Bandwidth 2Mbps
MAC layer protocol IEEE 802.11
Application type used CBR (Constant Bit Rate)
Mobility model used Random Waypoint Model
TABLE 3.SIMUALTION PARAMETERS
The following metrics is used to analyze the scalability
and performance of AODV by increasing the number
of nodes in the network from 100 to 1000 nodes. Five CBR
sessions are generated between randomly selected source-
destination pairs. Averages of five sessions are taken for
analysis.
The following three scenarios are considered for the
analysis.
IACSIT International Journal of Engineering and Technology, Vol.2, No.4, August 2010
ISSN: 1793-8236
398
1. Mobility is kept constant at a minimum speed of 0
m/s, a maximum speed of 10 m/s, and a pause time
of 30 m/s.
2. Varying the mobility speed from 10 m/s to 50 m/s
3. Varying the network load from 5 sessions to 30
sessions.
The following metrics are used to analyze the
performance of the proposed scheme.
A. Network Throughput
This value represents the ratio of the total number of
packets that reach their destination, to the total number of
packets sent by the source. It is calculates according to this
formula: Throughput = Packets Received / Packets Sent.
B. Average end-to-end delay of Data Packets
This is the average delay between the sending of the data
packet by the constant bit rate source and its receipt at the
corresponding constant bit rate receiver.
C. Routing overhead
Routing overhead is the total number of control packets
transmitted by nodes while establishing and maintaining
routes. Each hop-wise transmission of the control packet is
considered.
In order to evaluate and compare the performance of
proposed technique, a most widely used unipath on demand
protocol AODV is chosen.
Three scenarios are considered for the performance
evaluation.
1. Keeping the mobility of a node at a constant speed
2. Varying the mobility speed
3. Varying the network load
Scenario – I: Keeping the mobility of a node constant
Throughput variation with network size
15600
15800
16000
16200
16400
16600
16800
100 200 400 600 800 1000
Numbe r of Node s
Throughput (bps)
AODV
MORT
Fig. 2. Variation of throughput with network size.
Figure 2 shows the throughput comparison of MORT and
AODV. Packet delivery capacity of all these routing
techniques decreases as the number of nodes in the network
increases. This is due to the increasing number of route
breaks as the size of network increases. However, the
propsoed scheme outperforms AODV in packet delivery
capability for all sizes of network because most of the route
breaks are corrected with secondary paths at intermediate
nodes. This avoids packet drops at all the upstream nodes of
the intermediate node that detected the route break. On the
other hand, in AODV, all upstream nodes of the broken link
drop packets to the disconnected destinations as they do not
have secondary paths. Some of the packet drops are also
due to the congestion caused by high routing overhead in
AODV.
CTRL packets with network variation
0
1000
2000
3000
4000
5000
6000
100 200 400 600 800 1000
Numbe r of Nodes
Total Number of CTRL packets
(in s)
AODV
MORT
Fig. 3. Variation of routing overhead with network size.
Figure 3 shows the variation of routing overhead of two
routing techniques. The value increases with network size
because, the number of nodes communicating control
packets and number of route computations increase as the
network size increases. Number of route computations
increase with network size because of increase in number of
route breaks. AODV has higher routing overhead than
MORT at all network sizes. This is because, AODV
involves additional route computations and route error
packet transmission for recovering route breaks. Where as
in MORT route breaks can be resumed through the
secondary paths and only a limited number of route breaks
cause fresh route discoveries. Hence the proposed scheme
has lower routing overhead that of AODV.
Average Packet Transmission Delay
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
100 200 400 600 800 1000
Number of nodes
Packet Transmission Delay (in
s)
AODV
MORT
Fig. 4. Variation of packet transmission delay with network size.
Figure 4 shows the comparison of average packet
transmission delay experienced by data packets for AODV
and MORT. This metric reflects the delay involved in
resuming the sessions after route breaks have occurred. The
delay is high for AODV than MORT. But MORT has the
lowest delay value at all network sizes, as it finishes the
session with lowest number of route computations when
compared to AODV. The proposed scheme increases
throughput when compared to AODV. Reduction in routing
overhead enables MORT to scale to double the number of
nodes that AODV supports.
Scenario – II: Varying the mobility speed