Content uploaded by Benamar Kadri
Author content
All content in this area was uploaded by Benamar Kadri on Nov 11, 2014
Content may be subject to copyright.
Weight based DSR for Mobile Ad Hoc Networks
Benamar KADRI
STIC Lab., Department of
Telecommunications, University of
Tlemcen, Tlemcen, Algeria
benamarkadri@yahoo.fr
Mohammed FEHAM
STIC Lab., Department of
Telecommunications, University of
Tlemcen, Tlemcen, Algeria
m_feham@mail.univ-tlemcen.dz
Abdallah M’HAMED
National Institute of
Telecommunications, Evry, France
abdellah.mhamed@int.evry.edu
Abstract—Routing in ad hoc network is a great problematic,
since a good routing protocol must ensure fast and efficient
packet forwarding, which isn’t evident in ad hoc networks. In
literature there exists lot of routing protocols however they don’t
include all the aspects of ad hoc networks as mobility, device and
medium constraints which make these protocols not efficient for
some configuration and categories of ad hoc networks. Thus in
this paper we propose an improvement of Dynamic Source
Routing Protocol DSR in order to include some of the aspects of
ad hoc networks as mobility and energy by proposing a new
metric to evaluate route based on intermediate nodes weight
computed by combining the stability and the battery power of
nodes to choose the most stable and powered nodes for packet
forwarding. The paper is organized into five parts, in the first
ones we give an overview of ad hoc networks and routing
protocols; and a brief presentation of DSR. However the rest of
parts are consecrated for our proposed improvement for DSR
and simulation results of Weight Based Dynamic Source Routing
Protocol WDBSR finishing the paper by a conclusion and future
extension of this work.
Key words: MANETs, routing, energy, stability, weight,
WDBSR, DSR.
I. INTRODUCTION
obile ad hoc network (MANET) are a collection of
wireless mobile nodes forming a temporary network,
using as transmission medium radio waves. The used specter
for wireless transmissions is the specter situated around the
2.4 and around the 5 GHz [1]. The transmission range and the
emission power are regulated by laws in each country, ranging
from 10 m for Personal Area Networks to 100-200 m for
Local Area Networks [2].
Regarding its costless, facility of use and deployment,
MANET gets day after day new applications ranging from
military applications for connecting soldiers in battlefields and
civil or commercial application such as Public and Personal
Area Networks, other applications are recently under
development will also benefit from MANETs advantages such
as telemedicine, weather report and disaster environment such
as in seism. All these examples of use predict for some
envisioned MANETs to increase in size to reach the threshold
of thousands of nodes per system (commercial or military).
However in ad hoc network there is no concept of
centralized administration, to manage some tasks as security,
routing and others, therefore mobile nodes must collaborate
among themselves to accomplish these services. However due
to dynamic topology; the energy and the bandwidth constraints
due to the nature of devices and the transmission medium;
these tasks are not easily carried out. Thus any developed
protocol for ad hoc networks must take into consideration all
the aspects of ad hoc networks as mobility, energy and
bandwidth constraints to develop an efficient and effective
routing or security protocol [3].
In the remainder of this paper we are going to treat the
aspect of routing by proposing an improvement of Dynamic
Source Routing Protocol DSR to include some system
parameters as a new metric for route evaluation, trying in this
way to solve some of the problematic of ad hoc networks.
II. ROUTING IN AD HOC NETWORKS
In ad hoc networks, to ensure the delivery of a packet from
sender to destination, each node relay on its neighbouring
nodes to forward packets to nodes which are not in its
transmission range. Therefore any node in the network plays
two roles the first one as an ordinary node and the second one
as a router. The problematic in ad hoc network is the dynamic
topology of the network, resulting on a fast changing in routes
which must be efficiently handled by the underlying routing
protocol. Other problems exist due to the nature of nodes as
energy and computing power must also be carried out by the
routing protocol to ensure effective routing. Regarding the
technique and the strategy of routing we can classify them into
the following categories: reactive, proactive, and hybrid
protocols.
A. Reactive protocols
Under a reactive protocol, topology information is given
only when needed. Thus whenever a node wants to know the
route to a destination, it floods the network with a route
discovery request in order to get the sequence of nodes to the
destination [4]. This gives reduced average control traffic, and
an additional delay due to the fact that the route is not
immediately available. The much known reactive protocol is
DSR (Dynamic Source Routing)[4], however they exist other
like AODV (Ad hoc On-demand Distance Vector routing)[5].
These protocols have the inconvenient that they cause a
great congestion during the process of discovering routes;
however some protocols as DSR and DSDV have proven their
efficiency for ad hoc networks.
B. Proactive protocols
Proactive protocols are characterized by periodic exchange
of topology control messages, to update their routing tables.
Therefore, control traffic is more dense but constant, and
routes are instantly available. Some examples of these kind of
routing protocols are: OLSR (Optimized Link State
M
Routing)[6], OSPF (Open Shortest Path First) and FSR
(Fisheye State Routing)[7].
Theses protocols keep the network traffic within an
acceptable threshold due to the maintain of routing tables;
however when the topology changing is very frequent the
maintenance of routing table causes a great overhead due to the
number of exchanged data, making them not desired for high
mobility ad hoc networks.
C. Hybrid protocols
These protocols try to solve the problem of routing in ad
hoc networks by designing protocols having the advantages of
both reactive and proactive protocols; however there is no real
implementation of these protocols. Another problem is the
overhead due to the maintenance of the hierarchical
architecture.
Usually, the network is divided into regions called clusters,
and a node employs a proactive protocol for routing inside its
near neighbourhood’s region and a reactive protocol for
routing outside this region. Two known example of hybrid
routing protocols are ZRP (Zone Routing Protocol) [8] and
CBRP (Cluster Based Routing Protocol) [9].
In the remainder of this paper we focus our interest on
reactive protocols since these protocols have proven their
efficiency for ad hoc networks, especially on DSR since it’s
the most known and popular routing protocol, on which we are
going to apply our proposed metric to evaluate routes.
III. DYNAMIC SOURCE ROUTING PROTOCOL
The Dynamic Source Routing protocol (DSR) is a simple
and efficient routing protocol designed specifically for use in
multi-hop mobile ad hoc networks. Using DSR, the network is
completely self-organizing and self-configuring. The DSR
protocol can be described in the following points:
A. Route discovery:
This mechanism is launched whenever a node wishes to
send or contact a destination node which isn’t in its
transmission range; therefore it must obtain a route to that node
by launching the Route discovery mechanism. Normally the
sender must first search this route in its route cache if there is
no route it proceeds as follow:
-It creates a route request packets containing its address and
the address of the destination node; then it broadcast this
packet to all its neighbors using flooding.
-Each neighbor when receiving this request consults its
cache to find an eventual route to this destination to be
returned back to the sender; otherwise it rebroadcast the same
route request to all its neighbors after adding its address to the
header of the route request and learns from this request
information to be added to its cache. If the node has already
treated this route request it ignores the new received request by
verifying its sequence number since each route request is
identified by a unique sequence number.
The same procedure is executed by each neighboring node
until the route request arrives to destination which adds its
address at the end of the header and sends a route reply.
B. Route reply
This procedure is executed by a node after receiving a
route request destined to him, thus this node executes the
following actions:
-Adds this new route to its cache for future use.
-Adds it address at the end of the path contained in the
header of DSR packets.
-Replies to this request using unicast along the path
contained in the header.
C. Route maintenance:
when forwarding a packet each intermediate node is
responsible for confirming that the packet is correctly received
by the next node, however due to the dynamic topology and
the constraints of the wireless medium it may occur some
situation where a node doesn’t receive the acknowledgement of
reception from link layer of a given packet [4], therefore it
resends the same packet it until reaches a predefined value of
attempts [4]. Whenever this number of attempts was reached
this node consider this link as broken than it deletes each route
containing this link from its cache, than it generates a route
error packet to inform the source node and all intermediate
nodes about this link failure, in the same way each
intermediate node deletes all routes containing this route until
the route error packet arrives to its destination, which chooses
to launch a new route request or to find a new route in its route
cache.
D. Route cache
The route cache in DSR is used to maintain frequently
used routes in order to avoid new route discovery mechanism
which consumes lot of network resources, in the way that each
new discovered route is saved in the route cache of the
corresponding node for future use, a node can also learns from
route request to adds new routes to its cache, it also learns from
route error packets to update its cache.
E. Metric
The metric used to evaluate (choose route) in DSR is
usually the number of hops between a source node and a
destination, which is the metric used for most of conventional
routing protocol, however this mechanism isn’t desired for ad
hoc networks regarding their characteristics (mobility, devices
constraints). Because it may exist a route with the minimum of
hops however it contains some less powered or not stable
nodes which may causes link failure and therefore route errors.
S
D
N1
N2
N7
N6
N4
N5
N3
Route request
Route request
Route reply
Route reply
Figure 1. Route discovery mechanism
F. Limitation of DSR
Regarding the specifications of DSR we can conclude the
following limitations:
- It doesn’t take into consideration the capacity of each
node as power computing, because nodes with less computing
power may slow the forwarding of data flows.
- It doesn’t take into consideration the battery power of the
intermediate nodes, because a node can’t forward packets if it
hasn’t the sufficient energy power which causes link failures.
- It doesn’t include the aspect of stability of nodes in order
to choose the most stable nodes as intermediate nodes, because
unstable nodes causes topology changing which may launches
lot of route errors and therefore new route requests.
- Only the first discovered route is used, and the others are
cached for future use.
-No security mechanism is defined for DSR.
To overcome some of these limitations we are going to
propose an improvement of DSR in which we are going to use
a new metric including some of the characteristics (stability,
energy) of ad hoc networks to evaluate routes.
IV. WEIGHT BASED DSR
As we have said DSR suffers from lot of limitations
presented above. In order to overcome these limitations we are
going to make some improvement to the original
implementation of DSR in the way that we are going to
propose a new metric to better evaluate routes:
A. Weight based metric
Usually the metric used to evaluate route is based on the
number of hops (nodes) between the source and the destination
as in conventional networks, however the method is not
representative in ad hoc networks, which has other aspects to
include in route evaluation. Since it may exist a shortest route,
however it includes a less powered node which can’t forward
efficiently packets, or no stable nodes causing quick link break
down, which cause link failure and therefore route error to treat
this failure causing new route discovery requests which
overheads the network.
Ad hoc networks are characterized by a dynamic topology,
bandwidth and battery power constraint. Thus any metric for
mobile ad hoc routing protocol must take these parameters into
consideration, in the way that the final metric must include the
aspects of node mobility, security, battery state and link
bandwidth.
To do so we are going to propose a weight based metric in
which we include two system parameters battery power,
stability combined in order to compute a global weight used as
metric to evaluate routes.
1. Stability: Stability of nodes in ad hoc networks can be
defined as the possibility of a given node to be as long as
possible within the same neighbourhood. To compute stability
we must keep a table of one hop neighbours which is
periodically scanned to determine nodes have gone far from
the corresponding node.
To implements this we have made an improvement to the
MAC layer [10]of each node, to include a method to
periodically compute the number of neighbors, and define the
number of absent node (nodes have left the neighborhood of
the corresponding node). A node is marked as absent if we
don’t receive any MAC layer packet from that node during a
given period; in this way we compute the number of absent
nodes which is used in following equation to compute stability:
Stability = ((total number of neighbors)t – (number of absent
nodes)t + θ) / (total number of neighbors)t
A node is marked as stable if it has the greatest value of ST.
2. Battery power: As well as stability the energy level of
nodes is an important issue when searching route in ad hoc
networks, because nodes with less battery power have small
transmission range which causes link failure or it can’t
forward packet for long time. Therefore we have included the
energy power of nodes in order to choose the route containing
the most powered nodes to avoid link failure due to energy.
Including the aspect of energy in route selection may
minimize route error due to energy, in the way that the chosen
route is the one containing the most powered nodes which can
support dense traffic forwarding for long time, and whenever a
new route is needed we choose always the most powered route
which may differ from the fist one which equilibrates the use
of routes and the energy of intermediate nodes. This aspect
may maximize the node lifetime and therefore the system
lifetime.
In order to include the aspect of energy in route selection
we have used the battery level (
Bl
) of intermediate nodes,
and we choose the route containing the most powered nodes
which having the highest battery level. In this way the route
may serve for long time.
S
D
N1
N2
N7
N6
N4
N5
N3
Route error
Node N3 is death
his battery is 0
Figure 3. The effect of node battery power on
packet forwarding
S
D
N1
N2
N7
N6
N4
N5
N3
Route error
No stable nodes
N2, N3
N3
Figure 2. The effect of node stability on packet
forwarding
B. Weight computing
As seen in the previous section we have defined two
criteria to be considered when evaluating routes on which we
add the number of hops, in order to choose as main route the
one having the maximum power, the most stable nodes and the
minimum of hops. To accomplish this we propose to compute
the weight of each route as the sum of all these criteria.
Therefore when a route request is received by a destination it
computes the route-weight of this route request and compares
it with other route-weights and chooses the one with the
maximum weight.
To do so we have use a min-max strategy to evaluate and
choose routes:
1. Compute the node-weight (Nwi) of each node I contained in
the route as follow:
iii StBlNw
2. Compute the route-weight as the minimum of all node-
weights included in this route:
)( ir NwMinRw
,
ri
3. Choose as main route the one having the maximum route-
weight.
)( r
RwMaxMr
routes ofset theR /Rr
4. If two or more routes have the same route-weight or
whenever their route-weights are nearby ε
(
)( ji RwRw
) we choose as route the one with the
minimum of hops, (ε is a predefined value).
C. Implementation
To make in practice these specifications we have made the
following modifications:
1. Route request: it is similar to the original route request;
however in our case each intermediate node when receiving a
route request it computes its weight and inserts it in the route
request, to be used by the destination for route evaluation.
2. Route reply: the route reply packet is generated by the
destination node in response of a route request, however in our
case it differs from the original route reply, in the way that the
destination node doesn’t response immediately to route request
however it waits a predefined delay d to receive more route
requests then it compute for each received route request its
route-weight as defined in section IV.B. After the expiration of
the delay it replies by the route having the maximum route-
weight.
3. The route maintenance: no modifications are done on
the original route maintenance mechanism.
4. The route cache: we have used the same idea based on
the maximization of route-weights to choose the best routes.
We have also add a mechanism to periodically refresh the
cache in the way that any cached route when it exceeds in the
cache this period, it is automatically dropped and new route
request is launched, this aspect is included to always get fresh
routes which reflects best the state of the network (nodes
battery power and stability).
V. SIMULATION RESULTS
We investigate the performance of WBDSR by using the
ns-2 simulator [11] which is considered as the most powerful
and effective tool to test the performance of network protocols
for both conventional and wireless networks by giving all
possibilities to test all possible scenarios. We have compared
the original version of DSR with our proposed version
WBDSR to prove the utility of our improvement.
To do so we have used the CMU wireless extension, with
some modification in order to implement all the concepts
devoted above. The major modifications done on the standard
NS [11] implementation are done on the 802.11 MAC [10]
layer in order to compute the stability of nodes, we have also
made modification on the DSR implementation [4] to
implement the new route request mechanism described above.
In the way that each node when receiving a route request it
adds its weight computed as the sum of its energy level and
stability computed by the MAC layer. We have also modified
the route cache by adding some procedure to periodically
refresh the cache. Therefore any cached route which exceeds in
the cache this period is automatically dropped and new route
request is launched this aspect is included to always get fresh
route reflecting the state of the path, we have also injected four
CBR (Constant Bit Rate) connections with a packet length of
512 bytes, starting at different time during the simulation, the
remainder of the parameters of simulation are listed in table 1.
Table 1: Simulation parameters
Parameters
Values
Network size
670*670 m2
Number of CBR connections
4
Number of Nodes
25, 50
Max speed
20 m/s
Pause Time
0,60,120,300 s
Cache refresh
5 s
The value of ε
2
Wait time for route request d
0.25 s
Simulation time
200 s
S
D
N1
N2
N7
N6
N4
N5
N3
Route request
Route request
Route reply
N8
N9
Replies by the
best route
Route request
Figure 4. Weight based DSR
050 100 150 200 250 300
50
100
150
200
250
300
350
400
450
500
550
600
650
700
Route errors
WBDSR
DSR
Pause time (s)
Figure 5. Route errors with 25 nodes.
050 100 150 200 250 300
200
400
600
800
1000
1200
1400
1600
Route errors
WBDSR
DSR
Pause Time (s)
Figure 6. Route errors with 50 nodes.
In Figure 5 and 6 we have tested the performance of
WBDSR compared to the original DSR according to nodes
pause time, as we can observe in the two scenarios (with 25
and 50 nodes), WBDSR gives less route errors which may
improve the performance of the network because the
maintenance of route errors needs to launch new route requests
which causes a great overhead due to the flooding mechanism
used to broadcast route requests over the entire network, which
can block the network traffic in some situations and consume
lot of energy from nodes battery.
40 60 80 100 120 140 160 180 200
0,01
0,1
1
Delay (s)
Time (s)
WBDSR
DSR
Figure 7. End to End Delay with 25 nodes
20 40 60 80 100 120 140 160 180 200
1E-3
0,01
0,1
1
Delay (s)
Time (s)
WBDSR
DSR
Figure 8. End to End Delay with 50 nodes
As devoted in the previous simulations WBDSR may
improve the network performance, which is proved in the two
scenarios shown in Figure 7 and Figure 8 (with 25 and 50
nodes) in which WBDSR gives always best results compared
to the original DSR. The performance improvement is clearer
when the number of nodes get high, since WBDSR gives
always less delay for packet forwarding.
020 40 60 80 100
106
107
108
109
110
111
112
113
114
Nodes
WBDSR
DSR
Time (s)
Figure 9. system life time according to the number of nodes.
In Figure 9 we have fixed all the parameters and we have
varied the number of nodes from 10 nodes to 100 nodes and
we have measured the system life time as we can observe the
WBDSR always gives the longest system life time especially
in dense network when the number of nodes gets high.
050 100 150 200
106,0
106,5
107,0
107,5
108,0
108,5
109,0
109,5
110,0
110,5
111,0
111,5
112,0
112,5
113,0
Pause time (s)
Time (s)
WBDSR
DSR
Figure 10. End to End Delay with 50 nodes
In Figure 10 we have fixed the number of nodes to 50
nodes and we have varied the pause time from 0 s to 200 s
(static network), as it is shown WBDSR gives always the
longest system life time in both high mobile networks and
static network, because it periodically change the used route
with another one which equilibrate the use of the nodes which
increases the system life time.
VI. CONCLUSION
In this paper we have presented an improvement of
dynamic source routing protocol by proposing a new metric to
evaluate routes. This metric is based on nodes weight
computed by combining two parameters which are the power
of node and its stability assumed to be the most important
parameters in choosing routes. Then using these weight we can
choose the best route which may be the long one however it’s
the best regarding our two proposed arguments; whenever two
routes have near values of weights we choose the one with the
minimum of hops.
We have implemented this improvement in ns [11] which is
assumed to be the most effective tool to implement networks
simulation. Then using simulation in different situation we
have proven the efficiency of our improvements.
In the future we are going to extend the parameters
included to compute weight, by including security and a
mechanism to measure the computing power of intermediate
nodes including in this way all the characteristics of mobile ad
hoc networks in the route evaluation.
REFERENCE
[1] Houda Labiod, Hossam Afifi. De Bluetooth a WIFI securité, qualité
de service et aspects pratiques, hermes-science,2004.
[2]Guy Pujolle. Securité Wifi. Eyrolles, 2004.
[3] B.KADRI, A.M’HAMED, M. FEHAM. ―A new management scheme
of cluster based PKI for ad hoc networks using multi-signature‖.
IEEE Global Information Infrastructure Symposium, 2007.
[4]D. B. Johnson and D. A. Maltz, The Dynamic Source Routing
Protocol for Mobile Ad Hoc Networks (DSR), IETF Internet draft,
19 July 2004, http://www.ietf.org/internet-drafts/draft-ietf-manet-
dsr-10.txt
[5]C. E. Perkins et al. Ad hoc On-Demand Distance Vector (AODV)
Routing, RFC 3561, July 2003, http://www.ietf.org/rfc/rfc3561.txt.
[6]T. Clausen, and P. Jacquet, Optimized Link State Routing Protocol
(OLSR), RFC 3626, October 2003, http://ietf.org/rfc/rfc3626.txt
[7]Patroklos g. Argyroudis and Donal O’mahony. Secure Routing For
Mobile Ad Hoc Networks. IEEE Communications Surveys &
Tutorials • Third Quarter 2005.
[8]Haas, Z.J., Pearlman, M.R. and Samar, P. Zone Routing
Protocol(ZRP). IETF Internet Draft, draft-ietf-manet-zrp-04.txt,
July 2002.
[9]Mingliang Jiang, Jinyang Li, Yong Chiang Tay. Cluster Based
Routing Protocol(CBRP) Functional Specification. IETF Internet
Draft, draft-ietf-manet-cbrp-spec-00.txt August 1998.
[10]IEEE Computer Society LAN MAN Standards Committee. Wireless
LAN Medium Access Control (MAC) and Physical Layer (PHY)
Specifications, IEEE Std 802.11-1997.
[11]The Network simulator The Network Simulator ns-2. Project web
page available at http://www.isi.edu/nsnam/ns/.
