A Novel Routing Algorithm for Inter-Group Load
Balancing in Wireless Mesh Networks
Fakir Mashuque Alamgir 1
Assistant Professor
Department of Electrical and Electronic Engineering
East West University
Dhaka, Bangladesh
fma@ewubd.edu
Faisal Ahmed 2
Graduate Teaching Assistant
Department of Electrical and Electronic Engineering,
East West University
Dhaka, Bangladesh
ahmedfaisal065@gmail.com
Abstract — Wireless networks throughout the year have
experienced colossal innovative progressions. Because of
reconciliation of each of the three kinds of administrations viz.
voice, information, video, amid change in remote advances,
Bandwidth has turned into the most pivotal system asset that
should be upgraded for these bandwidth hungry administrations.
In multi-hop situations (in WMNs), execution relies upon the
routing protocols to appropriately pick courses, given the present
system conditions. To determine this issue, this paper proposes a
load balancing routing algorithm, called LBRA, in WMNs to
disperse the activity stack among different gateways. Through
simulation, we demonstrate that the proposed algorithm can
disperse the traffic load consistently among various gateways and
stifle the expansion in the routing path length at most 20%
contrasted with the shortest path routing.
Keywords—Load balancing; multiple gateways; routing
algorithm; wireless mesh networks.
I. INTRODUCTION
Wireless mesh networks (WMNs) are very popular in
wireless world now a days. Because of former communication
system that is connected point to point with other devices [1].
Mesh networks can pass on information through routing
technique. In this method, the information is sent to the
destination by hopping from one node to the next and so forth.
To provide a cost effective and dynamic high-bandwidth
978-1-5386-4110-1/18/$31.00 ©2018 IEEE
Mamun Miah 3
Department of Electrical and Electronic Engineering
East West University
Dhaka, Bangladesh
mamunmiah034@gmail.com
Hossain Mohammad 4
Department of Electrical and Electronic Engineering
East West University
Dhaka, Bangladesh
hossainmohammad021@gmail.com
Shourov Barua 5
Department of Electrical and Electronic Engineering
East West University
Dhaka, Bangladesh
shourovbarua5@gmail.com
network over a particular coverage area wireless mesh
network is one of the best infrastructures. Unlike WLAN
access points; this network is constructed of peer radio devices
which need not to be cabled. The wireless mesh network is
different from the traditional wired access network as data is
transported over multiple wireless hops. But this way of
transportation increases the collision, packet corruption and
other phenomena. As a result, gateway connection can get
congested and become a bottleneck for the entire network. To
lighten this problem we have to balance the gateway load of
wireless mesh router. In this article some load balancing
routing approaches are considered ([2]-[6]) and we will
discuss about the new multi-hop architecture called Wireless
Mesh Networks (WMNs), the gateway load balancing and will
show a novel routing algorithm for the WMNs to distribute
traffic loads.
II. ALGORITHM OF ROUTING
In this section, we initially characterize all gateways load
balance factor (GLBF) and after that present the load
balancing routing algorithm in wireless mesh networks.
A. All Gateways Load Balance Factor
To dominate the mobile nodes for discover an ultimate
routing path to the gateway to circulate the traffic load
uniformly among different gateways, we characterize the idea
of all gateways load balancing factor (GLBF) as takes after.
Without the loss of all extrapolation, we accept that the Table 1 shows the load balancing routing algorithm
network is associated. Else, either some mobile nodes are not (LBRA) in wireless mesh networks (WMNs). We show the
associated with any gateways, or the issue can be partitioned overview of this algorithm as follows.
into some sub-issues which can be settled freely.
STEP 1: For all vi (n  i  m+1), calculate the shortest path
B. Load Balancing Routing Algorithm in Wireless Mesh from vi to its nearest gateway using shortest path routing
Networks algorithm. Store vi in Vs, record the routing path in Pi,gi and
During the procedure of figuring the routing path to the compute its length in vector lengthi. Then we have to compute
gateways to accomplish load balancing among different the remaining available bandwidth of each gateway and
gateways, the length of the routing path computed by our gateway load factor. After that, we have to calculate all
calculation might be too long to compare with the one gateways load balance factor stored by LBF0.
ascertained by the shortest path routing algorithm. Reference STEP 2: For all vię Vs, repeat the following steps.
[7] demonstrates that throughput execution of wireless multi- STEP 3: Find all vj such that li,ję E and push vj into Stack.
hop session is soberly downgraded with increment of the STEP 4: If the Stack is empty, then returns to STEP 2.
quantity of hops. Along these lines, when our algorithm Otherwise, pop up the top element of the stack and assign this
computes a routing path to a further gateway, throughput element to vk. If k  m, then go to STEP 4-1. Otherwise, go to
execution may yet not be enhanced. In extraordinary instance STEP 4-2.
of discovering a directing way to a far further gateway, STEP 4-1: Save the current value of the gateway gi and gk.
throughput execution through this gateway might be reduced Recalculate the remaining available capacity of the gateway gi
genuinely and our calculation may bring more inferior and gk and the all gateways load balance factor stored by
throughput. So, we apply the shortest path routing algorithm LBFk. Then determine whether the following two conditions
as a pattern calculation and present a routing path increase are satisfied at the same time:
threshold to confine the length of routing path figured by our (1) LBFk < LBF0, 1+|Pkgk|  Lengthi+hthres. If they are
calculation in a sensible range. satisfied, then replace Pi,gi with {li,k}ĤPkgk and assign LBFk
to LBF0, go to STEP 4. Otherwise, restore the value of the
TABLE I. Load Balancing Routing Algorithm in Wireless Mesh Networks gateway gi and gk and go to STEP 4.
Load Balancing Routing Algorithm
STEP 4-2: Find all the link lk,lę E and push vl into the Stack.
(LBRA)
Then go to STEP 4.
Input: Graph G, hthres
Output: Pi,gi : m+1  i  n
Shortest Path Algorithm; LBRA exhaustively searches for all the possible routing
Store the corresponding information in Vs, Pi,gi, lengthi; paths from vi to vgi. So, the worst-case time complexity of this
for (all vg (m  g  1)) { algorithm occurs with a full connected network topology [9].
Calculate_Remaining_Bandwidth(vg); In this case, each mobile node has m links between the
Calculate_Load_Factor(LFg); gateways and (n-m-1) links between the other mobile nodes.
Calculate_GLBF(LBF0); Therefore, (n-m) mobile nodes search for ((n -m) m + (n-m-
} 1)!) paths. The shortest path routing (SPR) uses Dijkstra’s
for (all vię Vs ) { algorithm [8] for calculating the shortest path, the time
Push vj into Stack such that li,ję E; complexity of obtaining the path between a mobile node and
while (!IsEmpty(Stack)) { the gateway is O (n2). In SPR, each mobile node repeats this
vk = Get_top(); Pop_up();
operation m times. Then it determines the path to the gateway
if (k  m) {
Save(vgi , vgk);
within O(n). So, the time complexity of SRP is O(n2). Then
Recalculate(vgi , vgk); the time complexity of LBRA is the sum of ((n-m) m + (n-m-
Calculate(LBFk); 1)!) and O(n2). In the end, the time complexity of LBRA is
if ((LBFk < LBF0)&&( 1+|Pkgk|  O((n-m)!).
Lengthi+hthres)) { LBF0 = LBFk ;
Ĥ
Pi,gi = {li,k} Pk,gk;
} III. PERFORMANCE EVALUATION
else {
In this section, through simulation using ns-2.34 network
Restore(vgi , vgk);
} simulator [10] with CMU wireless extensions, we demonstrate
} the effectiveness of the proposed algorithm in terms of the
else { load on the gateways and the path length.
Push vl into the Stack such that the link lk,lę E;
A. Simulation Setup
}
}//while The simulation area is 1200m*1200m. The node mobility
}//for speed we chosen in the simulation varies from 10 to 20 m/s.
The random waypoint model is used as the mobility model of
the mobile nodes. Constant Bit Rate (CBR) traffic sources
with different packet generation rates are used to model represents the routing path increment threshold is unlimited.
different network scenarios. The maximum transmission range Because of obtaining similar results in other cases with
of each node is 250 meters. Simulations are run for 600 s [9]. different numbers of gateways, we set hthres = 1 in the
In addition, the gateways are deployed as follows. For 2 following evaluation.
gateways, they are deployed in the diagonal of the square. For
4 gateways, they are deployed in the vertex of the square. For C. Load Balance on Gateways
8 gateways, they are deployed in the vertex of the square and
in the midpoint of the edge of the square. This is because the We deploy 2, 4 and 8 gateways respectively in the
numbers of gateway nodes and their locations are carefully connected hybrid network to investigate the impact of SPR
determined to maximize the throughput between WMN and and our LBRA algorithm on the load balance on gateways
the Internet. The simulation parameters are given according to when the number of mobile nodes varies. The results are
Table 2. shown in Figure 4. As earlier noted, the locations of the
TABLE II. Simulation Parameters gateways are carefully determined to maximize the throughput
Parameter Value between WMN and the Internet. From Figure 4, we can see
that the all gateways load balance factor of the two algorithms
Simulation Area 1200m× 1200m
become low as the number of mobile nodes increases. The
Wireless MAC Interface IEEE 802.11 MAC
Propagation Model Two Ray Ground
degree of reducing the all gateways load balance factor of
Transmission Range 250m
LBRA is more obvious. The LBRA can distribute the load
Gateway Nodes 2, 4, 8 among gateways more uniformly [9].
CBR Packet Size 512 bytes
Mobility Model Random Waypoint
MANET node speed 10-20 m/s
Simulation Time 600 s

B. Optimal Path Increment Threshold Setting

Under the premise of distributing traffic load evenly among
multiple gateways, LBRA suppresses the increase in the hop
count caused by adjusting the routing path up to hthres. The
optimal routing path increment threshold is investigated from
the viewpoints of both the load on the gateways and the path
length [9].

Figure 2. 2 Gateways.

Figure 1. Relationship between hthres and all gateways load balance factor.

Figure 1 illustrates the relationship between h thres and all

Figure 3. 4 Gateways.
gateways load balance factor when the number of mobile
nodes varies and the number of gateways is set to 4. As shown
in Figure 3, the all gateways load balance factor does not
change if hthres is over 1. The case that hthres equals to 30

Figure 4. 8 Gateways.
D. The Impact on Average Routing Path Length
Figure 5 depicts how the average routing path length
between a mobile node and a gateway varies according to the
number of gateways when the number of mobile nodes is 40.
Figure 6 shows how the average routing path length between
a mobile node and a gateway varies according to the number
of mobile nodes when the number of gateways is 4. We find
the average routing path length of LBRA is at most 20%
higher than that of SPR. So, when using our LBRA to
distribute the load evenly among multiple gateways, the
increase of the routing path length is not large. This effect is
achieved by the introduction of the routing path increment
threshold hthres [9].
Figure 5. The impact of SPR and LBRA on average routing path length with
different number of gateways.
Figure 6. The impact of SPR and LBRA on average routing path length with
different number of mobile nodes.
IV. CONCLUSIONS
In this paper, we proposed a novel routing algorithm for
inter-group load balancing in wireless mesh networks to
distribute the traffic load on gateways. To avert the routing
path becoming too elongate, a routing path proliferation
threshold is proposed to confine the routing path in a
reasonably pasture. Through simulations, we showed that the
proposed algorithm LBRA can circulate the traffic loads
among multiple gateways in wireless mesh networks (WMNs)
invariantly and can circumscribe the proliferation of the
routing path to, at most, 20% compared to SPR. In the future,
we will inquire how to set the threshold dynamically
proceeding for nonidentical network scenarios.
ACKNOWLEDGEMENT
We would like to be so thankful to our creator for providing
us the mental and physical strength to conduct the research
works and letting us successfully complete the entire research.
Indeed, we have advanced a considerable measure of exertion
here. However, we have very frequently sought help from
many individuals and organizations. We thankfully
acknowledge their enormous contributions. We are also
thankful to our family members and our friends for their
constant and silent mental support and encouragement.
REFERENCES
[1] S. Corson and J. Macker, Mobile Ad hoc Networking (MANET):
Routing Protocol Performance issues and Evaluation Considerations,
Network Working Group, RFC2501, January 1999.
[2] Q. Le-Trung, P.E. Engelstad, T. Skeie and A. Taherkordi, Load-
Balance of Intra/Inter-MANET Traffic over Multiple Internet
Gateways, Proceedings of the 6th International Conference on
Advances in Mobile Computing and Multimedia, Pages 50-57, 2008.
[3] Vinicius C. M. Borges, Erik Dimitrov, Marilia Curado and Edmundo
Monteiro, RAILoB - A Routing Algorithm for Inter-cluster Load
Balancing in Wireless Mesh Networks. The 9th Annual IEEE
Consumer Communications and Networking Conference - Wireless
Consumer Communication and Networking. pp: 904-909, Las Vegas,
NV, USA, 2012.
[4] Choi, Hyoung-Gyu, SSeung-Jae Han, Load Balancing Routing for
Wireless Mesh Networks: An Adaptive Partitioning Approach. 5th
IEEE Consumer Communications and Networking Conference, Las
Vegas, NV, USA 2008. pp:1-5
[5] Wen Song, Xuming Fang, Routing with Congestion Control and Load
Balancing in Wireless Mesh Networks, the 6th International
Conference on ITS Telecommunications Proceedings, 2006
[6] Chungui Liu, Yantai Shu, Lianfang Zhang, Jun Li, Efficient Multiple
Gateways Load-Balancing and QoS Routing in Wireless Mesh
Networks. 4th International Conference on Wireless Communications,
Networking and Mobile Computing, Dalian, China. 2008 pp.1-4.
[7] J. Li, C. Blake, D. S. J. D. Couto, H. Lee and R. Morris, Capacity of Ad
Hoc Wireless Networks, in Proc. ACM MobiCom '01, pp.61-69, July
2001.
[8] E. W. Dijkstra. A note on two problems in connection with graphs.
Numerische Mathematik, 1(6):269-270, 1959.
[9] Y. Yan, L. Ci, R. Zhang, Z. Wang, "Load balancing routing algorithm
among multiple gateways in MANET with Internet connectivity", Proc.
16th Int. Conf. Advanced Communication Technology, pp. 388-392, 16-
19 Feb. 2014.
[10] The Network Simulator – NS-2, http://www.isi.edu/nsnam/ns/