performance improvement of optimized link state...
TRANSCRIPT
Performance Improvement of
Optimized Link State Routing (OLSR) Protocol
Navaid Akhter1, Ammar Masood
2, Irfan Laone
3
Institute of Avionics and Aeronautics/Department of Avionics, Air University, Islamabad, Pakistan 1,2,3
Abstract- OLSR, a leading proactive protocol of MANET
maintains consistent and up to date network topology at
all the times and has emerged as the choice for MANETs
due to low latency for route determination. Hence, OLSR
generate a large amount of control overhead in order to
maintain an up-to-date routing table which consumes
bandwidth that should have been employed by user data
traffic instead. This paper addresses this issue by
optimizing OLSR under specific network and mobility
conditions which are actually more of practical interest
and thereby, our work does have a valuable contribution
to provide guidelines for large number of cases of general
interest. The proposal has shown to consistently
outperform the default implementation by reducing the
routing overhead under specific network and mobility
conditions considered at no extra cost. Other parameters
like data traffic and end-to-end delay also improved with
the approach presented in this study which shows the
efficiency of the scheme selected.
Keywords: MANETs, Routing Protocols, OLSR,
Improvement in Control Messages’ Intervals, Optimality in
performance
1 Introduction
Research concerning MANETs is currently of great
interest. The performance of MANET is related to the
efficiency of the routing protocols in adapting to frequently
changing network topology and link status [1]. Because of
the importance of routing protocols in the dynamic multi
hop networks, a number of routing protocols have been
proposed in the last few years; concurrently, a great deal of
research work is being undertaken by researchers to
improve their performances. In OLSR (a leading MANET
routing protocol), maintaining an up-to-date routing table
for the entire network calls for excessive communication
between the nodes as periodic control messages updates
are flooded throughout the network. Hence OLSR generate
a large amount of control overhead which consumes
valuable bandwidth that should have been employed by
user data traffic instead. Therefore, excessive control
overhead in OLSR is detrimental to its overall performance
in data forwarding, which has been analyzed for
improvement in our research work. Other parameters like
data traffic and end-to-end delay also improved with the
approach presented in this study.
2 MANET Routing Protocols
Mobile Adhoc Network (MANET) is an
autonomous system of mobile nodes connected by wireless
links [2]. MANET routing protocols are based on how
routing information is acquired and maintained by the
mobile nodes and thus, can be divided into proactive,
reactive and hybrid routing protocols [3]. With proactive
routing protocol, nodes in a MANET continuously evaluate
routes to all the reachable nodes and attempt to maintain
consistent, up-to-date routing information. On the other
hand, in reactive routing protocols for MANETs (also
called “on-demand” routing protocols), routing paths are
explored only when needed. Hybrid routing protocols are
proposed to combine the merits of both proactive and
reactive routing protocols and to overcome their
shortcomings.
3 Optimized Link State Routing
(OLSR) Protocol
The Optimized Link State Routing Protocol (OLSR)
[4] is developed for MANETs and does not need central
administrative system to handle its routing process.
Because of its proactive characteristic, the protocol
provides all the routing information to all participating
hosts in the network at all times. However, as a drawback,
OLSR protocol needs that each host periodically send the
updated topology information throughout the entire
network by flooding. This increases the protocol’s
bandwidth usage as the routing overhead is high. Although,
flooding in OLSR is minimized by the Multi Point Relays
(MPRs), which are the only nodes allowed to forward the
topological messages [5,6], still the routing overhead is
high as compared to reactive routing protocols.
3.1 Control Messages Intervals
OLSR employs two types of control messages:
Hello messages and TC messages.
3.1.1 Hello Interval
This parameter represents the frequency of
generating a Hello message. Hello Interval determines the
time between successive Hello messages, which is set to 2
seconds by default. Hello messages are never forwarded.
3.1.2 TC Interval
This parameter represents the frequency of
generating a TC message. In OLSR, the rate of the
topological state updates is the sending rate of TC
messages. TC messages are broadcasted periodically
within the TC interval, to other MPRs, which can further
relay the information to further MPRs. TC messages are
broadcasted once per refreshing period and the default
value is 5 seconds. TC messages are one of the major
sources of overhead in OLSR, as they are flooded
throughout the network, but they are essential to maintain
consistent connectivity knowledge of complete network.
3.2 Problem in OLSR
One advantage of OLSR is that it provides lower
route discovery latency than on-demand protocols because
of its proactive nature. But the flip side is that OLSR
generates a large amount of control overhead which
consumes precious bandwidth. Since the resources in
wireless networks are severely constrained, the increased
channel contention could lead to network congestion
resulting in significant lowering of network performance.
Further, scalability issues arise in OLSR due to the
excessive routing message overhead caused by the
increased network population. The size of routing table
grows non-linearly with the increase in number of nodes
and the control messages can block the actual data packets.
Hence, excessive control overhead in OLSR is detrimental
to its overall performance in data forwarding and poses a
research challenge that need to be addressed.
3.3 Proposed solution
Optimization of local and global topology
dissemination intervals (i.e. Hello and TC intervals
respectively) is proposed under specific network and
mobility conditions which are actually more of practical
interest and thereby, our work does have a valuable
contribution to provide guidelines for large number of
cases of general interest that result in low routing overhead
(as compared to the default settings) and thus beneficial for
OLSR performance. This study targets on reduction in
control overhead with improvement in performance of
OLSR by optimizing control messages intervals.
3.3.1 Logical Reasoning of proposed solution
Hello messages are broadcasted periodically for
link sensing and neighbor detection. This is also required
to complete the MPR selection process. After the MPR
selection process is completed, TC messages are generated
and are disseminated throughout the network. Subsequent
to the receipt of these TC messages, the nodes calculate the
routing table and the links are available for data
communication. Further, MPRs broadcast the TC messages
in the network to maintain a consistent and up-to-date view
of complete network topology. In case of any topology
changes, the MPR selection process is re-initiated and the
routing table is re-computed by the nodes [8]. MANETs
require minimum control overhead to reduce channel
contention and battery consumption problems. TC
messages share a large amount of overhead in OLSR
because of its global dissemination nature. Decreasing the
broadcast frequency of TC messages reduces the overall
routing traffic sent while not incurring any degradation in
throughput / end-to-end delay under specific network and
mobility conditions as shown in this study. Further, due to
frequent topology changes caused by high mobility, the
routing information needs to be updated more frequently so
as to update the topology changes and guarantee the
correctness of route selection. This requires that nodes of
OLSR employed MANET detect link changes more
quickly and broadcast topology updates with lesser delay.
This can be achieved by increasing the Hello messages
sending rate for faster response to the link and neighbor
changes (especially in case of high mobility scenarios);
hence, providing better throughput as compared to the
default Hello interval. The increase in routing overhead
because of the increase in Hello sending rate above is
compensated by the reduction in routing overhead due to
the decrease in TC messages sending rate as mentioned
earlier.
4 Related Work
The authors of OLSR in RFC 3626 (OLSR) [4]
pointed out that the nodes may send control messages at
different rates, if beneficial for specific deployment. Many
strategies have been proposed by OLSR researchers using
different performance metrics to improve the performance
of OLSR by varying control messages intervals [7, 8, 9, 10,
11, 12, and 13]. However, these works usually target to
reduce control overhead while having certain deficiencies
and implementation complexities. Our work, however,
does not include any added complexity or depends upon
any measurement of network parameters and provides
improved performance of OLSR under specific network
and mobility conditions by just modifying the OLSR
control messages intervals. With our approach, now
network is able to achieve an increase in data traffic
received (vis-à-vis the payload with default control
messages intervals) while routing traffic and end-to-end
delay are both reduced. We have constructed fairly robust
scenarios for experiments to investigate the effect of
control messages intervals on the routing overhead of
OLSR.
5 Performance Evaluation
Because of the unavailability of wide range of real
MANETs, the performance analysis of wireless
applications or protocols in the context of MANETs often
require to be evaluated through simulation studies [14].
The performance analysis on a real network (if available)
can be rather tedious if large networks are considered
(typically hundreds of nodes). This is why simulation is an
important tool in the sense that it can often help to improve
or validate protocols [15]. OPNet Modeler 14.5 network
simulator was used for analysing the performance of OLSR
in this study.
5.1 Choice of Network and Mobility
Conditions
The MANET routing protocols perform
differently under different network & environmental
factors like node mobility, number of nodes, number of
source-destination pairs, traffic type, traffic intensity,
propagation models etc. For the purpose of performance
analysis in this study, we selected two factors: Node
mobility and Number of nodes, because of their major
impact on the mechanics of the protocol vis-à-vis routing
overhead. We started with a carefully designed network
scenario for all the experiments and varied one parameter
at a time and thus stressed the network in different axis as
shown in Table 1.
Table 1
Network & Mobility Factors
As mobility has the most significant impact on
MANET routing protocols [16], scenarios have been
constructed to evaluate the proposed solution against
varying nodes speed (from 5 mps to 40 mps) while keeping
the number of nodes fixed. The speed range has been
selected with keeping in view the speed selected for
extreme practical scenarios (low to high) and most of the
other research work in this area. Table 2 depicts the
parameters selected for the Scenarios.
Table 2
SIMULATION PARAMETERS
5.2 Performance Metrics
The performance metrics investigated during this
study were the data traffic received and routing overhead in
OLSR protocol vis-à-vis its improvement by optimizing the
OLSR control messages interval under various network
and mobility conditions. However, end-to-end packets
delay was also kept under-check so as to ensure that the
optimization of control messages for improvement in
routing overhead do not degrade this parameter. The
definition of improved performance is that the routing
protocol must provide applications with high data traffic
received, minimal routing overhead and low end-to-end
delay.
6 Results
6.1 Experiments
The simulation study adopts a step-by-step
performance optimization approach. Firstly, the impact of
each control messages’ interval of OLSR has been
analyzed distinctly on the data traffic received, routing
traffic overhead and the end-to-end packets delay by
stressing the mobility factor. The outcome of the results
has been analyzed further to see how these control
messages’ interval can be optimized simultaneously so as
to efficiently maximize the data traffic received (payload)
while minimizing the routing overhead and end-to-end
packets delay. The steps are depicted in Figure 1.
Figure 1: Simulation study steps
6.2 Experiment No 1
In experiment No 1, the Hello message update
rate has been increased from default value (2 seconds) to
various values like 1.9, 1.8, 1.7, 1.6, 1.5 seconds etc in
order to facilitate the routing protocol to speed up the
adaptation to neighbor changes while keeping the TC
interval at 5 seconds (default value). The value of the state
holding timer interval (neighbor hold time) was adjusted
accordingly. After various iterations, it was found that
Hello interval of 1.8 seconds (TC at default value)
provides the best balance between data traffic and routing
traffic and the results are shown in Figure 2.
Figure 2: Results of data traffic and routing traffic vs speed with change
in Hello interval
In Figure 2, the data traffic and routing traffic are
plotted on Y-axis and variation in speed is plotted on X-
axis. In all the simulations across specific range of nodes
speed while changing the Hello interval and keeping the
TC interval fixed, it can be appreciated that increase in
Hello sending rate (i.e. Hello 1.8 secs) from the default
value (Hello 2 secs) improves the data traffic received as it
helps the routing protocol to quickly adapts the changes in
neighbors and update the routing tables accordingly. On
the other hand, the routing traffic overhead also increases
with the increase in Hello sending rate which clearly depict
that although fast Hello messages improves the protocol
reactivity to link failures; however this is at the cost of
increased routing overhead.
Hence, from Experiment No 1, it is concluded that an
improvement in data traffic received by increasing the
Hello messages sending rate is at the expense of increased
routing traffic overhead.
6.3 Experiment No 2
In experiment No 2, the TC interval has been
decreased from default value (5 seconds) to various values
like 6, 7, 7.5 seconds etc in order to reduce the routing
overhead while keeping the Hello interval at 2 seconds
(default value). The value of state holding timer interval
(topology hold time) was adjusted accordingly. Since the
Hello interval is kept constant, the reduction in overall
routing overhead is the result of decrease in TC messages
overhead. The TC interval of 7.5 seconds was found to be
the optimized value under the considered mobility
conditions that provide a decrease in routing traffic
overhead while having almost no effect on data traffic and
the results are as shown in Figure 3.
Figure 3: Results of data traffic and routing traffic vs Speed with change
in TC interval
In Figure 3, the routing traffic and data traffic are
plotted on Y-axis and variation in speed is plotted on X-
axis. Firstly observing the routing traffic behavior with
default values of OLSR, it is revealed that as the speed
increases, the routing traffic decreases. This is because of
the reason that as mobility increases, link breakages
increases and therefore TC messages are either not
generated or if they are generated than they are not
forwarded to the entire network. Now with the modified
settings, it is evident that as TC messages sending rate is
reduced from 5 seconds (default) to 7.5 seconds, the
routing traffic is less as compared to the routing traffic at
default TC interval. Further, it is observed that decreasing
the TC sending rates from default value to 7.5 secs
although reduces large routing overhead; brings no
significant change in data traffic received. This is because
of the fact that repetitive TC messages are broadcasted
throughout the network to maintain the network topology.
Lowering down the sending rate of these TC messages
under specific mobility conditions although reduces large
routing overhead, brings no significant change in data
traffic received at the nodes which is more sensitive to the
change in Hello interval than the TC interval.
6.4 Outcome of Experiment No 1 and
Experiment No 2
Through simulations it has been explored that TC
messages generate more overhead than Hello messages
because TC messages are forwarded globally to each node
in the network while Hello messages are only exchanged
locally between neighboring nodes. Increasing the Hello
messages sending rate helps the routing protocol to quickly
adapt the changes in neighbors and update the routing
tables accordingly. Hello interval rate has been increased
from default value (2 secs) to 1.8 secs in order to speed up
the adaptation to neighbor changes and thus achieving
higher data traffic received than what it is achieved at the
default value. This is particularly to cater the declined
performance of OLSR under high mobility scenarios.
Decreasing TC messages sending rate leads to significant
reduction in control overhead but do not downgrade the
data traffic received under specific mobility conditions
considered in the experiments.
6.5 Experiment No 3
Experiments 1 and 2 provides a comprehensive
understanding of OLSR’s control timers’ behavior vis-à-vis
performance metrics and gives insightful guidance in
optimizing these timers for an improved performance in
data traffic received while introducing low routing
overhead as compared to the default values. The results
have been exploited further in experiment 3 to formulate
that how these two timers can be optimized simultaneously
under considered network and mobility conditions so as to
efficiently minimize the routing overhead while achieving
maximum data traffic received without compromising on
end-to-end delay. The OLSR control messages intervals
Hello 1.8 secs and TC 7.5 secs (as discussed in
experiments 1 and 2) were selected and compared against
the default intervals to see if there is any improvement as
stated above.
6.5.1 Routing traffic vs Speed:
In Figure 4, the routing traffic is plotted on Y-
axis and variation in speed is plotted on X-axis.
Figure 4: Result of Routing traffic vs Speed with change in HELLO and
TC intervals
First observing the routing traffic behavior with
default values of OLSR (i.e. Hello def_TC def), it is
revealed that as speed increases the routing traffic
decreases. This is because of the reason that as mobility
increases, link breakages increases and therefore TC
messages are either not generated or if they are generated
than they are not forwarded to the entire network. This
results into decrease in routing traffic as the speed
increases and vice versa. Now with the modified intervals
(i.e. Hello 1.8_TC 7.5), the similar behavior of decrease in
routing overhead with increase in mobility is observed as
stated above. Further, as the TC messages sending rate is
reduced from 5 seconds (default) to 7.5 seconds, the
routing traffic is noticeably reduced as compared to the
routing traffic at default TC interval. Also due to the
increase in Hello sending rate from 2 seconds (default) to
1.8 seconds, the routing traffic would have increased (as
observed in experiment 1). However, this has been
compensated with the reduction of large routing overhead
due to the decrease in TC interval. Hence, the overall result
is the reduction of routing overhead as compared to the
routing overhead with default Hello and TC values. The
average reduction in routing overhead achieved with the
modified Hello and TC intervals is 14.06 %.
6.5.2 Data traffic vs Speed
In Figure 5, the data traffic is plotted on Y-axis
and variation in speed is plotted on X-axis.
Figure 5: Result of Data traffic vs Speed with change in HELLO and
TC intervals
Firstly, observing the data traffic behavior with default
values of OLSR (i.e. Hello def_TC def); it is revealed that
as speed increases the data traffic decreases. This is
because of the reason that as mobility increases, link
breakages increases and therefore the nodes are unable to
forward the data to the required destination which resulted
into dropping of data packets before reaching to the
destinations. Now with the modified interval i.e. Hello
1.8_TC 7.5, similar behavior of decrease in data traffic is
observed with the increase in node’s mobility (due to the
same reason as mentioned above). However the data traffic
is now improved than what it is achieved with the default
OLSR intervals (i.e. Hello def_TC def). This is because of
the increase in Hello messages sending rate which speed up
the routing protocol’s adaptation to neighbor changes and
route maintenance and thus resulting into less data drop
and increase in data traffic received at the destinations.
Further, it is observed that both the curves are tending to
converge at very low speeds and at very high speeds. It is
because of the reason that at very low speeds, there are no
significant changes in neighbors so the default interval as
well as modified Hello interval works almost the same
manner and the change in Hello interval does not make any
difference. Similarly at very high speeds, the topology
changes might be too dynamic to be captured by the
periodic updates of OLSR with default as well as with the
modified settings so the change in Hello interval does not
make any significant impact in this regime also. The
average increase in data traffic achieved with the modified
values of Hello and TC intervals is 6.19%
6.5.3 End to End packets delay vs Speed:
In Figure 6, the delay is plotted on Y-axis and
variation in speed is plotted on X-axis. Firstly, observing
the end to end packets delay behavior with default values
of OLSR (i.e. Hello def_TC def); it is revealed that as
speed increases, the end to end packets delay decreases.
Figure 6: Result of End to End packets delay vs Speed with change in
HELLO and TC intervals
This is because of the reason that as mobility increases,
link breakages increases and therefore less number of
source-destination pairs are now available at high speeds as
compared to the scenarios at low speeds. This results into
increase in channel capacity because of the occupation of
same number of available channels now with less number
of source destination pairs. Hence, the packets reach to the
destination with lesser problems of channel contention and
therefore end to end packets delay decreases. The similar
behavior is observed with the modified intervals of OLSR
because of the same reason as mentioned above. However,
now with the modified intervals, the end to end packets
delay is less as compared to the default settings. This is
because of the increase in hello sending rate which
increases the routing protocol’s adaptation to neighbor
changes and route maintenance that decreases the overall
end to end packets delay.
6.6 Summary of Experiment No 3
The simulation results demonstrated that by
optimizing the Hello and TC intervals, optimality in the
routing protocol performance is achieved under specific
mobility factors considered. Through simulations it has
been explored that increasing rate of hello update leads to
improvement in link establishment and node status
maintenance. Further, decreasing rate of TC updates leads
to significant reduction in control overhead but do not
downgrade the data traffic received under specific mobility
conditions considered. Hello interval has been slightly
decreased from default value (2 secs) to 1.8 secs in order to
alleviate the degraded performance of OLSR under high mobility scenarios thus achieving higher data traffic
received than the default value. Increase in routing traffic
due to increase in Hello interval has been compensated by
decreasing the TC sending rate from 5 to 7.5 secs which
drastically reduced the overall routing overhead while not
posing any significant impact on data traffic received. This
also resolves the problem of high routing overhead of
OLSR (generated due to its proactive nature) under the
specific mobility conditions considered. In the proposed
Hello and TC intervals, OLSR is now able to sustain an
increased data traffic received compared to the default
values of Hello and TC intervals and at the same time, both
the routing traffic and end-to-end packet delay are also
reduced.
7. Conclusion
MANET is an autonomous system of mobile
nodes connected by wireless links. The performance of
MANET is related to the efficiency of the routing
protocols. OLSR, a well known proactive protocol has
emerged as the choice for MANETs (especially for delay
sensitive applications) due to low latency for route
determination. But at the same, time associated high
routing overhead (due to proactive nature) has emerged as
a major performance issue in OLSR. In this study, we have
addressed this issue by optimizing the OLSR under specific
mobility conditions which are actually more of practical
interest and thereby, our work does have a valuable
contribution to provide guidelines for large number of
cases of general interest. The default parameters of Hello
and TC intervals of OLSR are selected such that the
network performance is improved. The behavior of the
routing protocol is tested based on the influence of node
mobility using various performance metrics. From the
results of simulations, it is concluded that the optimization
of OLSR control messages intervals has shown to
consistently outperform the default implementation of
OLSR under specific mobility conditions considered
during this study. We envisage undertaking research to
analyze the scalability of OLSR protocol vis-à-vis control
messages intervals with number of nodes and to set the
boundary limits through detailed simulation studies.
Furthermore, the performance analysis of OLSR protocol
must be analyzed with realistic mobility models [17] so as
to finalize realistic protocol performance.
8. References
[1] M. Abolhasan, T. Wysocki, and E. Dutkiewicz, A
review of routing protocols for mobile adhoc networks,
Elsevier Journal of Ad Hoc Networks, 2004.
[2] The Book of Visions 2000, Visions of the
Wireless World, IST – WSI Project, November 2000.
[3] S. Corson, J. Macker, MANET RFC 2501,
MANET: Routing Protocol Performance Issues and
Evaluation Characteristics, January 1999.
[4] C Adjih, T. Clausen, P. Jacquet, A. Laouiti, P.
Minet, P. Muhlethaler, A. Qayyum, L. Viennot: Optimized
Link State Routing Protocol, RFC3626, IETF, October
2003.
[5] A. Qayyum, L. Viennot, A. Laouiti: Multipoint
relaying: An efficient technique for flooding in mobile
wireless networks, INRIA research report N 3898, March
2000, INRIA Rocquencourt, France.
http://www.inria.fr/rrrt/rr-3898.html.
[6] Jerome Harri, Christian Bonnet and Fethi Filali,
OLSR and MPR: Mutual Dependences and Performances,
EURECOM research report RR-05-138, 2005.
[7] Yang Cheng Huang, Saleem Bhatti, Daryl Parker,
TUNING OLSR, The 17th Annual IEEE International
Symposium on Personal, Indoor and Mobile Radio
Communications, PIMRC’06.
[8] C. Gomez, D. Garcia, J. Paradells Improving
Performance of a Real Ad-hoc Network by Tuning OLSR
Parameters, 10th IEEE Symposium on Computers and
Communications, ISCC 2005.
[9] Carlos Miguel Tavares Calafate, Roman Garcia,
Pietro Manzoni, Optimizing the implementation of a
MANET routing protocol in a heterogeneous environment,
Proceedings of Eighth IEEE International Symposium on
Computers and Communications, ISCC '03.
[10] Mounir Benzaid, Pascale Minet, Khaldoun Al
Agha, Integrating fast mobility in the OLSR routing
protocol, Mobile and Wireless Communication Networks,
MWCN 2002.
[11] P. Samar and Z. Haas, Strategies for Broadcasting
Updates by Proactive Routing Protocols in Mobile Ad hoc
Networks, Proceedings of the IEEE Military
Communications Conference (MILCOM), Anaheim,
California, USA, October 2002.
[12] Pedro E. Villanueva-Peña, Thomas Kunz, Pramod
Dhakal, Extending Network Knowledge: Making OLSR a
Quality of Service Conducive Protocol, IWCMC 2006.
[13] F. Bai, N. Sadagopan, A. Helmy, IMPORTANT:
A framework to systematically analyze the Impact of
Mobility on Performance of Routing protocols for Adhoc
NeTworks, IEEE INFOCOM, 2003.
[14] S. Kurkowski, T. Camp, and M. Colagrosso,
"MANET Simulation Studies: The Incredibles," ACM
SIGMOBILE Mobile Computing and Communications
Review (MC2R), pp. 50-61, October, 2005.
[15] OPNet tutorial Modeling concepts reference
manual.
[16] S Gowrishankar , T G Basavaraju, S. K. Sarka,
Effect of Random Mobility Models Pattern in Mobile Ad
hoc Networks, International Journal of Computer Science
and Network Security, VOL.7 No.6, June 2007.
[17] Jungkeun Yoon, Mingyan Liu, Brian Noble,
Random Waypoint Considered Harmful, IEEE INFOCOM
2003.