Journal of Software Engineering and Applications, 2012, 5, 36-41
doi:10.4236/jsea.2012.512b008 Published Online December 2012 (http://www.SciRP.org/journal/jsea)
Benefit of Selecting Number of Active Mesh Routers in
Disaster Oriented Wireless Mesh Network
Panu Avakul1, Hiroki Nishiyama1, Nei Kato1, Yoshitaka Shimizu2, Tomoaki Kumagai2
1Graduate School of Information Sciences, Tohoku University, Sendai, Japan; 2NTT Network Innovation Laboratories, NTT Corpo-
ration, Yokosuka, Japan.
Received 2012
ABSTRACT
Communications is one of the most critical factors in disaster recovery process. However, after a major disaster, exist-
ing communications infrastructures may be heavily damaged or even completely unusable. It is necessary that commu-
nicationsare to be promptly restored to the disaster area, which is the goal of our national project. The project aims to
build three tiers wireless mesh network from remaining wireless access points in order to provide communications ser-
vices to the disaster area. This work introduces a unique multiple tiers wireless mesh network project. In addition, this
work also illustrates merits in optimizing the number of mesh routersin order to achieve the optimum performance by
presenting both theoretical and simulation results of a specific scenario of multiple tiers wireless mesh network.
Keywords: Wireless Mesh Network; Throughput; Interference; Collision; WiFi
1. Introduction
Following the 2011 Tohoku earthquake and tsunami, the
field of emergency network deployment has gainedmuch
attention from research institutions all over Japan. This is
because drastic disaster such as earthquake or tsunami-
will incapacitate the existing communications infrastruc-
ture. Therefore, it is important to be able to restore
communications in the disaster area using whatever
means necessary. In addition, it ismentioned by [1] that it
is more complicated to deploy a new communications
system where the previous system still exists. An ongo-
ing national project aims to restore communications
within the disaster area using the remaining resources,
namely the remaining wireless access points, to con-
structs a wireless mesh backbone and to provide commu-
nications to users in the affected area who are referred to
as Mesh Client (MC).
Wireless Mesh Network (WMN) is a type of infra-
structure where each participating node or a Mesh Router
(MR) both sends and relays information similar to an
ad-hoc network with the exception that MR does not
usually have power constraint or mobility problem. In
addition, many MRs are also likely to act as wireless
Access Point (AP) and provide communications service
to MCs within its area similar to that of an ordinary
wireless AP. WMN has been gaining attention from re-
search communities due to the low-cost and rapid de-
ployment, whichare suitable properties for an emergency
network role. In addition, it also inherits many other
valuable characteristics from ad-hoc network such de-
centralized design and distributed communications[2].
In addition to being able to relay information between
MRs, MR is also usually equipped with an addition radio
interface to acts as a wireless AP andto provide commu-
nications services to users within its area. This implies
that the MCs will be able to connect to a network like
connecting to WiFi hot spot, but the communication is
relayed through the wireless mesh backbone instead of
the traditional wired connection.Due to this unique char-
acteristic of this multiple tiers WMN, the number of ac-
tive MRs plays a significant role in determining the per-
formance of the whole network.
It was mentioned that our project relies extensively on
the remaining wireless APs that are already deployed in
an uncontrolled manner,and number of active MRs can
affect the performance in multiple-tiers of the pro-
ject.Therefore, in order to deploy the project successfully
with the optimum performance, we need to optimize the
number of MRs.This work briefly introduces the national
project and investigates the effect of using the optimum
number of MRs in a unique WMN project. We introduce
a theoretical performance of a specific scenario of multi-
ple-tier WMN, which is backed up by simulation results.
The remaining sections of this paper are divided into
following: Section 2introduces related works and dis-

Benefit of Selecting Number of Active Mesh Routers in Disaster Oriented Wireless Mesh Network
37
cusses the structure of the project. Section 3shows theo-
retical discussion of the scenario. Section 4 presents the
simulation results and discussion. Finally, Section 5 pre-
sents the conclusion.
2. Background
2.1. Related Works
Reference [3]presents Extreme Networking System (ENS)
architechture, which is a three tiers network that is very
similar to our project’s architechture. According to the
article, ENS was experimentally deployed on November
2005, at San Diego and many useful emergency response
statistics, such as traffic statistic, andperformance of the
architechture, was collected. The authors also present
many challenges in deploying wireless mesh network in
the case of emergency response. One of the major dif-
ference of ENS and our project is that our project plans
to rely on existing inflastruture such as remaining wire-
less APs. A more detail explaination regarding our pro-
ject will be given in Section 2.2.
In addition to ENS presented by [3], authors of [4] also
present another similar network architechture to our pro-
ject called the hybrid WMN. Hybrid WMN aims to en-
hances the performance of metapolian WMN with the
addition of wired APs. While hybrid WMN is not in-
tended for usage in an emergency response situation, it is
a great idea which may also help increase the overall
performance of emergency response WMN. Since in an
ideal situation, some remaining wireless APs may still
have wired connection available.
Reference[5] investigates the impact of inter-cell in-
terference on WLAN performance. The authors’ testbed
experiment shows that inter-cell inteference can greatly
reduce the WLAN performance under the TCP domi-
nated common office traffic pattern.
Reference [6] presents a novel gateway selection
method for multiple tiers WMN that focuses on deploy-
ment in disaster area. The method presented ultilize the
concept of Collision Domain (CD) to choose the most
suitable MR to act as Mesh Gateway (MG)in order to
have the best overall system throughput.
Reference [7] presents a deployment evaluation of the
Roofnet wireless mesh network, which is an unplanned
WMN. The author suggests that having a denser MR
may help improve the average throughput of the network
due to the posibility of chosing shorter high quality links
with the cost of having a higher average hop-count.
However, the author also shows that the performance
suffers from multi-hops transmission due to inter-hop
interference. In addtition, [8]shows that in a multi-hop
WMN there is fairness problem where MRs that has a
higher hop-count to the gateway achieve much lower
throughput than those with lower hop-count.
2.2. Project Structure and Characteristics
2.2.1. Three Tiers Network
As shown in Figure 1, the project is composed of several
components. The Moveable and Deployable Resource
Units (MDRU) acts as a gateway that provides connec-
tions from the system to outside network such as internet.
The MR is an ordinary wireless AP which is configured
to act as wireless mesh backbone in order to provide
communications service to the MCs. MG is a specific
MR that communicates directly with the MDRU. The
overall project can be partitions into three separated tiers:
network facility tier, mesh tier, and clients tier.
The network facility tier covers the communication
between the MDRU(s) and MGs, which is done in a spe-
cial 25GHz band. The mesh tier is the wireless mesh
backbone level where the MGs and MRs form a wireless
mesh backbone network using the 5GHz band. The MGs
act as mesh portal to bridge the communication between
the network facility tier and the mesh tier. Finally, the
clients tier is the level, which each MG and MR provide
communications service to MCs within its vicinity over
the 2.4GHz band. Since the links at the network facility
tier are assumed to be high performance links, this work
focus on the performance in mesh tier and clients tier.
2.2.2. Uniq ue Depl o yment
A scenario of how the project will be deployed is the
MDRU will be deploy to the disaster area prior to disas-
ter. The MDRUis capable of carrying a certain number of
MGs in order to have some control over the topology of
the system. After arrival, the MDRU will configure any
remaining wireless APs in the area into MRs or MGs and
restore communications service to MCs in the area.
2.2.3. Uncontrolled MRs Placement
Another major characteristic of the project is that since
MRs are just commonly deployed wireless APs, and it is
pointed out by [9] that common 802.11 APs are usually
deployed in an unplanned or random fashion. Therefore,
there is a good chance that the deployment will not be
optimized, and thus causing the drop in performance.
Figure 1.The structure of the project.

Benefit of Selecting Number of Active Mesh Routers in Disaster Oriented Wireless Mesh Network
38
3. Effect of Number of MRs
3.1. Effect of Number of MRs
CD is an important concept that can be used to estimate
the capacity of WMN as shown in [10]. Since wireless
links share the same medium, it is necessary that only
one device within the same interfering range should be
transmitting at any given time to have a successful
transmission. In another word, if a certain link n is active,
any other link within the same interfering range of link n
must be inactive in order for transmission at link n to be
successful. A CD is defined by [10] to be a set of links
that have to be inactive for a transmission at a certain
link to be successful including the transmitting link itself.
Figure 2 illustrates a chain topology of a two tiers WMN
where there are six MRs, which can only transmit to an
adjacent MR. The rightmost device is the MG that is as-
sumed to be the sink of all traffic. Each MR has to
transmit G amount of traffic from its own clients tier and
any other amount of traffic forwarded by its upstream
MR(s). For example, MR2 has to forward G amount of
traffic from its clients tier plus what MR1 forwarded
which is another G to the total of 2G amount of traffic. It
is assumed that the interfering range is two times the
transmission range; therefore, the CD of link between
MR4 and MR5 is a set of {(MR2, MR3),(MR3,
MR4),(MR4, MR5),(MR5, MR6),(MR6, MG)}.
The bottleneck collision domain (BCD) is defined by
[10] to be the CD that has to forward the most traffic. For
instance, the CD of link MR4 and MR5 has to forward
all traffic forwarded by each link within the set of its CD.
Therefore, CD of link MR4 and MR5 has to forward total
of 2G+3G+4G+5G+6G = 20G. The same calculation can
also be done on CD of other links, but in the case of sce-
nario in Figure 2,CD of link MR4 and MR5 has to for-
ward the most traffic. Hence, it is the BCD of this chain
topology. According to the previous calculation, CD of
link between MR4 and MR5 has to forward 20G amount
of traffic. However, the capacity is bound by the MAC
layer capacity C, thus can be represent by
Figure 2. A chain topology where each MR sends the same
amount of traffic to the MG.
20G

C. (1)
By solving inequality in Equation (1), we can see that
the maximum throughput available to each MR or max
G
is
max 20
C
G (2)
From the analysis above, we can see that the more
general case of Equation (2) is
max
C
G
A
MT
 (3)
whereAMT is the total amount of traffic forwarded by
BCD divided by G, which will be 20 in the case of CD of
link MR4 and MR5.
3.2. Considered Scenario
In order to show the effect of number of MRs, we con-
sider following scenario illustrates in Figure 3(a) where
there are one MG, three MRs and six MCs. Some proper-
ties of the scenario are as following:
● All MCs can connect to any available MRs.(All MCs
are within the communication range of any MRs.)
● Each MR operates in different channel in a way that
they do not have inter-cell interference introduced in
[5].
● The mesh tier links operate in one common channel.
● Each MRs will always have the same number of MCs
within its cell.
In addition, this scenario can be narrow down to three
sub-scenarios as follow:
Figure 3. (a) Topology for considered scenario; (b) Sub-
scenario 1, consisting of one active MR; (c) Sub-scenario 2,
consisting of two active MRs; (d) Sub-scenario 3, consisting
of three active MRs.

Benefit of Selecting Number of Active Mesh Routers in Disaster Oriented Wireless Mesh Network
39
1) Only MR1 is active and all MCs are associated with
MR1 as illustrated in Figure 3(b).
2) Either MR1 and MR2, or MR1 and MR3 are active.
In this sub-scenario, each MR will have to serve three
MCs as illustrated in Figure 3(c).
3) All MRs are active and each has to serve two MCs
as illustrated in Figure 3(d).
It is possible to use the concept of BCD to estimate the
maximum throughput in both the mesh tier and the cli-
ents tier in each of these sub-scenarios so that we can see
the effect of number of MRs in this scenario.
3.3. Sub-scenario 1
In the clients tier, MR1 serves all six MCs as wireless AP.
Since all devices used only one channel within a given
cell, we can easily concluded that the CD of each link is
the same set that contain all links; thus all links are BCD
(all links need to transmitted same amount of traffic.)By
using Equation (3) whereCis the MAC layer capacity of
the clients tier or c C and AMTequals to 6, because there
are six MC each offering 1G traffic.We can concluded
that max G (throughput available to each MC) of clients
tier or c C is
6
c
c
C
G (4)
In the mesh tier, there is only MR1 and the MG.
However, the performance should be calculated based on
throughput available to each MC rather than throughput
available to each MR. Therefore, the max G of mesh tier-
should be divided by number of MCs served by each MR
or 6 in this case to formulate m G of
6
m
m
C
G (5)
where m G the MAC layer capacity of mesh tier.
3.4. Sub-scenario 2
In this sub-scenario, each MR has toserve three MC.
Therefore, using Equation (3) and AMT equal to 3 results
in c
C of
G
c
C
c
3
(6)
However, in the mesh tier there are two operating MR.
We can clearly see from Figure 3(c) that CD of both
links in the sub-scenario are BCN and each CD has to
carry 1G + 2G = 3G amount of traffic. By using Equation
(3) and AMT equals to 3 and the fact that each MR serves
three MCs, the resulting m G is
9
m
m
C
G (7)
3.5. Sub-scenario 3
The clients tier of this sub-scenario has three total active
APs and each has to serve two MCs. Using similar
analysis to sub-scenario 1 and sub-scenario 2 in previous
sections, we can see that AMT is 2 thus resulting with
2
c
c
C
G (8)
In the mesh tier, similar to sub-scenario 2, we can see
from Figure 3(d) that CD of any link includes all links,
and thus are BCNs. In this sub-scenario any BCNs has to
carry 1G+1G+3G=5G (AMT=5) amount of traffic and
that each MR serves 2 MCs, hence, resulting with
10
m
m
C
G (9)
3.6. Scenario Discussion
Since our project is a multiple tiers WMN, the perform-
ance of the project will be restricted by the tier with
lower performance. We can clearly see from Table 1 or
Figure 4 that the performance in the mesh tier decreases
with increasing number of MR. This is because of the
interference modeled by BCD concepts. On the other
hand, the performance in the clients tier benefits from the
extra capacity of additional MR operating in nonoverlap-
ping channel. Therefore, it is important to optimize this
trade off, which results from the number of activating
MRsin the area, in order to provide the best maximum
throughput to each MC.
All metrics are summarized in Table 1 whileFigure 4
shows plots of m
G and c C with different value of m G
for different 802.11a data rates (Gm-54 for 54Mbps,
Gm-24 for 24Mbp, and Gm-12 for 12Mbps). Each value
of Cfor both c
C and m G are experimentally deter-
mined by simulations for each data rate. According to
Figure 4, the maximum throughput available to each MC
is bounded by either m
G or c C. Since m G and c C
rely on each other, the lower value of the two will be the
upper bound of the maximum throughput. Therefore, by
using this concept, it is possible to estimate the optimum
number of MRs that will give the maximum average
throughput to each MC. We can easily see from Figure 4
Table 1. Summary of all values of this section.
Metrics
Sub-scenario Gm Gc
1
C
m
6
C
c
6
2
C
m
9
C
c
3
3
C
m
10
C
c
2

Benefit of Selecting Number of Active Mesh Routers in Disaster Oriented Wireless Mesh Network
40
Figure 4. Theoretical value of Gm and Gc for different data
rate.
that the optimum value of MR used for 802.11a with
54Mbps data rate is three MRs.This is because the capac-
ity in the mesh tier is large enough to accommodate the
extra capacity in the clients tier, which results from hav-
ing three APs operate in nonoverlapping channel.
4. Simulation Results and Discussion
Simulations are conducted to confirm this effect dis-
cussed in section 3. The simulation scenarios are set up
in Qualnet 5.1 with topologies similar to that of Figure
3(b), (c) and (d). Each MR is set up to operate as AP in
different nonoverlappingchannel in the clients tier;thus,
interference between each MR will not exist within the
clients tier. In additional to clients tier interface, each
MR also has another interface for communicating in the
mesh tier. In mesh tier, all MRs operate in one common
channel, which means the collision domains contain all
links. Each MC sends CBR traffic to MG and thus all
traffic only originate from MCs and flow to MG.
Figure 5 shows results of the simulation scenarios
where the line graphs are the theoretical results from
Figure 4,and the bar graphs show the simulation results
of each mesh tier’s data rate (54 Mbps in blue, 24 Mbps
in redand 12 Mbps in green.) In the single MR scenario,
the simulated results represent the fact that the average
throughputs are bounded by c C, which has lower value
than all m G of all data rate. The results of two operating
MRs scenario differ from that of the single operating MR
scenario in that the average throughput per MC of
12Mbps case should now be bounded by m G with data
rate of 12Mbps (in green).This is because the result of
m
G with data rate of 12Mbps is now lower than that of
c
C as shown in Figure 5. However, since the average
throughput per MC of c C is still lower than those of
m
G of 24Mbps and 54Mbps, the throughput per MC of
those two cases are still bounded by c
C. Finally, in the
Figure 5. Simulation results and theoretical results for dif-
ferent data rate.
case of 3 active MRs, the theoretical results show that the
simulation results should be bounded by m
G of each
case. As shown in Figure 5, the simulation results clearly
follow those of the theoretical results. This shows that
the theoretical results are accurate, and we can concluded
that we can determine the number of active MR(s),
which would yield optimum number of MRs for each
m
G (of different data rate). From Figure 5, we can de-
duct that the most optimized number of MRs in m
G
equals to 54Mbps, 24Mbps, and 12Mbps are 3 MRs, 2 or
3 MRs, and 2 MR respectively. This is because these
combinations give the best per MC performance.
5. Conclusion
Previous sections show that in order for the project to
achieves the optimum performance, the set of active MRs
will need to be determined to give the best possible per-
formance. In the future, we aim to apply this idea into a
more general case of multiple tiers WMN. In addition,
many other important factors such as channel assignment
in the mesh tier, and inter-cell interference present in [5]
should also be considered when deciding optimum set of
active MRs.
In this paper, we briefly explain the national project
“The R&D on the reconfigurable communication re-
source unit for disaster recovery,” which is a unique mul-
tiple tiers WMN. We point out that by optimizing the
number of active MRs, we can achieve the optimum
performanceby using the concept of BCD and the unique
characteristic of multiple tiers WMN. In addition, we
also show both theoretical and simulation performance
based on a specific scenario of multiple tiers WMN. Our
results showed that by choosing the right number or
combination of active MRs, we can optimize the per-
formance of the project.

Benefit of Selecting Number of Active Mesh Routers in Disaster Oriented Wireless Mesh Network
41
6. Acknowledgements
This work was conducted under the national project,
“The R&D on the reconfigurable communication re-
source unit for disaster recovery”, supported by the Min-
istry of Internal Affairs and Communications (MIC),
Japan.
REFERENCES
[1] B.S. Manoj and Alexandra Hubenko Baker, “Communi-
cation challenges in emergency response,” Communica-
tions of the ACM – Emergency response information sys-
tems: emerging trends and technologies, Vol. 50, No. 3,
2007, pp. 51-53. doi:10.1145/1226736.1226765
[2] P. H. Pathak and R. Dutta, “A Survey of Network Design
Problems and Joint Design Approaches in Wireless Mesh
Networks,” IEEE Communications Surveys & Tutorials,
Vol. 13, No. 3, 2010, pp. 396-428.
doi:10.1109/SURV.2011.060710.00062
[3] B. Braunstein, T. Trimble, R. Mishra, B. S. Manoj, L.
Lenert and R. R. Rao, “Challenges in Using of Distrib-
uted Wireless Mesh Networks,”Proceedings of the 3rd In-
ternational ISCRAM Conference, Newark, NJ, May 2006
[4] W. Fu and D. P. Agrawal, “Capacity of Hybrid Wireless
Mesh Network with Random APs,”IEEE Transactions on
Mobile Computing, Vol. 10, No. 12, 2011.
doi:10.1109/TMC.2011.247
[5] M. A.Ergin, K. Ramachandran and M.Gruteser, “Under-
standing the effect of access point density on wireless
LAN performance,”Proceedings of the 13th annual ACM
international conference on Mobile computing and net-
working MobiCom, Montreal, 9-14 September 2007, pp.
350-353. doi:10.1145/1287853.1287902
[6] W. Liu, H. Nishiyama, N. Kato, Y. Shimizu and T. Ku-
magai, “A Novel Gateway Selection Method to Maximize
the System Throughput of Wireless Mesh Network De-
ployed in Disaster Areas,”23rd IEEE International Sym-
posium on Personal, Indoor and Mobile Radio Commu-
nication PIMRC, Sydney, 9-12 September 2012.
[7] J. Bicket, D. Aguayo, S. Biswas and R. Morris, “Archi-
tecture and evaluation of an unplanned 802.11b mesh
network,”MobiCom Proceeding of the 11th annual inter-
national conference on Mobile computing and networking,
Cologne, August 28- September 2 2005. pp. 31-42.
doi:10.1145/1080829.1080833
[8] J. Camp, J. Robinson, C. Steger and E. Knightly, “Meas-
urement Driven Deployment of a Two-Tier Urban Mesh
Access Network,”MobiSys Proceeding of the 4th interna-
tional conference on Mobile systems, applications and
services, Uppsala, 19-22 June 2006. pp. 96-109.
doi:10.1145/1134680.1134691
[9] A. Akella, G. Judd, S. Seshan and P. Steenkiste,
“Self-management in chaotic wireless deploy-
ments,”MobiCom Proceedings of the 11th annual interna-
tional conference on Mobile computing and networking,
Cologne, August 28 - September 2 2005.
[10] J. Jun and M. L. Sichitiu, “The Nominal Capacity of
Wireless Mesh Networks,” IEEE Wireless Communica-
tions, Vol. 10, No. 5, 2003, pp. 8-14.
doi:10.1109/MWC.2003.1241089