APPLYING LOAD BALANCE TO REDUCE THE ENERGY

CONSUMPTION OF THE VIRTUAL ROUTING PROTOCOL

Angelo Bannack and Luiz Carlos Pessoa Albini

NR2/LARSIS − Department of Informatics, Federal University of Paran´a, Curitiba, Brazil

Keywords:

Ad hoc networks, Routing, Energy consumption, Load balance.

Abstract:

Energy consumption is a key issue in mobile ad hoc networks.To reduce the energy consumption of the network

interface, it is necessary to turn it off, immediately affecting the routing protocols. Therefore, routing protocols

and energy saving algorithms must be designed to work within each other. This article presents the use of

energy management, using the load balance technique with single path, over the Virtual Routing Protocol

(VRP). VRP is a hybrid routing protocol that can achieve a high delivery ratio, though it has a high energy

consumption. Using load balance, it is possible to minimize this drawback, while keeping the high delivery

ratio from VRP. It is shown that the basic VRP consumes more energy than AODV and DSR. While using the

load balance technique, the energy consumption of VRP becomes very similar to AODV and DSR.

1 INTRODUCTION

The Virtual Routing Protocol (VRP) (Albini et al.,

2006) is a hybrid routing protocol for MANETs,

which uses a virtual structure to route packages. The

virtual structure deﬁnes the proactive part of the pro-

tocol, i.e. a connection from unit u to unit v, in the

virtual structure, means that unit u must maintain an

updated route to reach unit v. Unit u is called a scout,

and each scout is responsible for maintaining routes

to a small subset of other units (called peers). Figure

1 exempliﬁes a virtual structure for VRP.

a+1

a−1

Units peered by w

Scouts for w

s−1 s−1

Figure 1: Virtual Structure Example.

The use of the virtual structure makes VRP able

to maintain a high delivery ratio (Albini et al., 2006).

Another protocol, called Virtual Distance Vector

(VDV) (Robba and Maestrini, 2007), also employs a

virtual structure, but the goal of VDV is to reduce the

delay to build a route between any source-destination

pair. Though both protocol use the same virtual

structure concept, they are quite different: VRP is

based on source routing and achieves a high delivery

rate, while VDV is based on distance vectors and

achieve a small delay to build routes. However, they

are very similar in one point: they use more energy

to achieve their goals when compared to AODV

(Perkins and Royer, 1999) and DSR (Johnson and

Maltz, 1996), the two major distance vector and

source routing based routing protocols for MANETs.

Figure 2 compares the network lifetime for AODV,

DSR and VRP.

Energy consumption is critical in

MANETs (Feeney and Nilsson, 2001), thus sav-

ing energy is vital to maintain the network alive

as long as possible. It is possible to reduce the

energy consumption of hardware components, such

as CPU, display, discs and network interface. It is

also possible to use some algorithms to turn off the

network interface or even keep it in a low energy

consumption state most of the time (Cho et al., 2005;

Hundewale et al., 2007; Iqbal et al., 2006). Even

though turning off the network interface increases

the single unit lifetime, it has a direct impact on the

routing protocol. Thus, energy aware algorithms and

routing protocols must be designed to work within

each other (Feeney and Nilsson, 2001).

One possible technique that can be applied on

routing protocols to increase the network lifetime

is the load balance (Cho et al., 2005; Hundewale

et al., 2007; Iqbal et al., 2006). Load balance does

Bannack A. and Carlos Pessoa Albini L. (2009).

APPLYING LOAD BALANCE TO REDUCE THE ENERGY CONSUMPTION OF THE VIRTUAL ROUTING PROTOCOL.

In Proceedings of the International Conference on Wireless Information Networks and Systems, pages 32-37

DOI: 10.5220/0002236300320037

 SciTePress

not aim to reduce the total energy consumption, but

to uniformly distribute it among the entire network.

It has two variations, multipath and single path

protocols. In multipath load balance, also known

as Multi-Path Routing (Ziane and Mellouk, 2005;

Lee and Gerla, 2001), data messages follow into

distinct routes simultaneously from the source to the

destination. The multi-path technique can achieve

a better load balancing than shortest path routing

if the number of paths in use is high (Ganjali and

Keshavarzian, 2004). The single path load balance

technique consists in ﬁnding a good route to send

data messages from the source to the destination

and using it until a better route is found. The great

challenge of such technique is to decide when a route

is not good anymore and a new route must be found

and used.

This article describes the application of the single

path load balance technique on the Virtual Routing

Protocol to reduce its major drawback, the excessive

energy consumption. Using such technique on VRP,

it is possible to reduce its energy consumption,

making it very similar to those of DSR and AODV,

without loosing its main advantage, the high delivery

ratio. All results were obtained through simulations,

using GloMoSim and considering the energy con-

sumption model presented in (Feeney and Nilsson,

2001).

Figure 2: Network Lifetime.

The rest of this article is organized as follows:

Section 2 details the application of the load bal-

ance technique over the Virtual Routing Protocol;

Section 3 describes the modiﬁcations made on Glo-

MoSim to simulate energy consumption; Section 4

contains the simulation results and Section 5 draws

the conclusions and future work.

2 VRP WITH LOAD BALANCE

To use the load balance technique on VRP, it is nec-

essary to change the virtual path construction phase

of the VRP. It is essential that VRP chooses the vir-

tual path with the highest residual energy. To do so,

all scouts must know, or at least estimate, the residual

energy of the physical routes to their peers.

The residual energy of a physical route from unit

i to unit j, denoted by re

i, j

, is the minimum of the

remaining energy of all units in the route. Consider

Vre

i, j

= {e

= e

,...,e

= e

} as the vector of

remaining energy of all units in the route from i to

j, then re

i, j

= min{e

,...,e

n−1

}. Note that the re-

maining energy of the source i and the destination j

are not considered.

To compute re

u,v

between any scout-peer pair, the

Route Reply messages must be changed including a

ﬁeld called route energy. When such a message tra-

verses the network, it collects the residual energy of

all units in the route. Thus, when it arrives at the scout

u, it knows the residual energy of the route connect-

ing itself with peer v. This information is maintained

within the virtual structure as weights on the edges

connecting u with its peers. It is important to empha-

size that weights are kept locally.

Recalling that the main change on VRP is on the

virtual path construction, now the virtual path be-

tween the source and the destination is not built en-

tirely by the source. To build a route from a source

(unit s) to a destination (unit d) in the new routing

protocol, unit s follows the following steps:

1. If d is a physical neighbor of s, then the route is

trivial and k = d.

2. Otherwise, if s is a scout to d, then k = d.

3. Otherwise:

(a) Computes the distance between itself and d in

the virtual structure, D

s,d

(b) Finds the unit k such that k ∈ P

and ∀m ∈

i,k

> re

i,m

, where P

is the set of peers of

unit s.

k,d

. If D

k,d

≤ D

s,d

then chooses k

as the next virtual hop. If D

k,d

> D

s,d

, return to

step 3(b) and choose another k.

(d) If s does not have an up-to-date route to k, then

return to step 3.b and choose another k.

(e) If there is no k that satisﬁes steps 3(b) and 3(c),

disconsider step 3.b and choose a unit k which

satisﬁes steps 3(a) and 3(c) only.

After ﬁnding unit k, s sends the message to k

through the physical route proactively maintained by

itself. When unit k receives the message, it veriﬁes if

it is the destination, if so the protocol ends. If unit k is

not the destination, it repeats the abovesteps consider-

ing itself as the origin, s = k. This process is repeated

until the message arrives at the destination. Note that

all units within the physical route between s and k be-

have exactly in the same way as they do in VRP. Re-

APPLYING LOAD BALANCE TO REDUCE THE ENERGY CONSUMPTION OF THE VIRTUAL ROUTING

PROTOCOL

calling that VRP uses source routing, i.e., data mes-

sages carry the entire route they must traverse, units

in the physical route between s and k just forward the

data message to the next physical hop.

3 ENERGY CONSUMPTION ON

GLOMOSIM

The Global Mobile Information System Simulator Li-

brary (GloMoSim) is a simulation environment to

wireless networks. In this work, GloMoSim was

adapted to include the energy consumption function-

ality and to stop the message redirection when the bat-

tery of the unit depletes. All parameters depicted here

are from (Bannack and Albini, 2008).

The energy consumption of each unit is not only

related with the network interface, other hardwares

like CPU, display or memory contribute to it (Feeney

and Nilsson, 2001). When a message is transmitted,

it is necessary to process and even store it, consuming

energy. Without loss of generality, these values are

grouped in a constant ∆. Further, the average energy

consumed to keep the network interface in idle mode

is called δ.

The energy used to send and receive a message

can be split in the following parameters:

• Send Preparation Phase (Σ): energy used dur-

ing send message preparation phase. It includes

the encapsulation, the inter-frame times, chang-

ing mode on the network interface communica-

tion, and sending the 802.11 MAC preamble;

• Send Message (σ): energy spent to send one byte

of the message, headers and data;

• Receive Preparation Phase (Ω): energy used dur-

ing receive message preparation phase. It includes

the decapsulation, the inter-frame times, chang-

ing mode on the network interface communica-

tion, and receiving the 802.11 MAC preamble;

• Receive Message (ω): energy spent to receive one

byte of the message, headers and data.

After each simulated second, the residual energy

of the unit (γ) is decreased by energy consumption

value to keep the node and the NIC alive (if there

were no transmission / reception on the previous sec-

ond): γ

t+1

= γ

− (∆ + δ). If there were a message

transmission on the previous simulated second, the

residual energy of the unit is decreased by (where b

is the number of bytes sent in this second): γ

t+1

− (∆+ Σ + (b∗ σ)). If there were a message recep-

tion on the previous simulated second, it is decreased

by (where b

′

is the number of bytes received in this

second): γ

t+1

= γ

− (∆ + Ω+ (b

′

∗ ω)). If γ

t+1

= 0,

the unit is considered unachievable(turned off)by any

other unit of the network, and it will not send or re-

ceive any message.

The energy consumption model implemented is

linear, i.e., the energy used to transmit/receive a mes-

sage depends only on the message size. When a node

does not have sufﬁcient energy to completely send

or receive a message, its battery is emptied and the

message discarded. The remain energy, before being

emptied, is added to the energy used to keep the node

alive.

Table 1 (Bannack and Albini, 2008) show the val-

ues for each of the above variables to send and receive

messages, respectively. These values were obtained

using the real energy consumption values of a Com-

paq WL110 connected to a HP IPAQ 3600.

Table 1: Energy used to send and receive messages.

Parameter Energy (pWh)

Σ 44,777

Ω 30,749

∆ 161,507,937

σ 161

ω 111

δ 113,055,556

4 SIMULATION RESULTS

Simulations were made using GloMoSim 2.03 with

the modiﬁcations speciﬁed in Section 3 and in (Ban-

nack and Albini, 2008). All simulations were per-

formed on an Intel Pentium Xeon 3.2GHz, 4GB

of RAM, running Debian Etch 4.1.1-21 with kernel

2.6.18-4-686. The network lifetime was measured

considering the time from the network initialization

until the ﬁrst unit runs out of battery.

4.1 Scenarios

Three different scenarios were evaluated: varying the

network density, the maximum speed of the units

and the throughput. The common way to vary the

network density is to vary the number of units in

a constant area, though it is possible to obtain the

same results maintaining the number of units con-

stant and varying the network dimentions. Without

any impact on the results, the network density was

varied by increasing the network dimentions from

50m

to 100.000.000m

, while maintaining the num-

ber of units and their transmission range constant.

The maximum speed of the units was varied from

0m/s to 20m/s and the throughput from 1024bps to

WINSYS 2009 - International Conference on Wireless Information Networks and Systems

512.000bps for each sender. Besides, all simulations

consider the parameters presented in table 2. It is im-

portant to point out that the cbr trafﬁc considers 10

sources, sending data packages of 512 bytes to other

10 units, and data packages are sent from the start of

the simulation until the ﬁrst unit of the network runs

out of energy.

Table 2: Parameters used in the simulations.

Parameter Value

UNITS 50

RAIO TRANSMITION RANGE 250m

TRAFFIC CBR

SIZE OF DATA PACKAGES 512 bytes

SIMULATION-TIME 14H

NODE-PLACEMENT uniform

MOBILITY random waypoint

MOBILITY-WP-PAUSE 0

MOBILITY-WP-MIN-SPEED 0

MOBILITY-POS-GRANULAR 0.5

PROPAGATION-LIMIT -111.0 dBm

PROPAGATION-PATHLOSS free-space

NOISE-FIGURE 10.0 dBm

TEMPERATURE 300.0 K

RADIO-TYPE radio-accnoise

RADIO-FREQUENCY 2.4e9Hz

RADIO-BANDWIDTH 11,000,000 bps

RADIO-RX-TYPE SNR-bounded

RADIO-RX-SNR-THRESHOLD 10.0 dB

RADIO-TX-POWER 7.005 dBm

RADIO-ANTENNA-GAIN 0.0 dB

RADIO-RX-SENSITIVITY -91.0 dBm

RADIO-RX-THRESHOLD -81.0 dBm

MAC-PROTOCOL 802.11

PROMISCUOUS-MODE yes

NETWORK-PROTOCOL IP

NETWORK-OUTPUT-QUEUE 100

4.2 Results

Figures 3, 4 and 5 show the simulation results vary-

ing the network density. In these simulations the

maximum speed of the units is set to 20m/s and the

throughput to 16384bps. In ﬁgure 3, it is possible

to notice that the network lifetime of VRP with load

balance is very symilar to the lifetime of AODV and

DSR, while the network lifetime when using VRP is

much smaller. Further,the energy necessary to deliver

a single data package (Figure 4) is much smaller when

VRP with load balance is compared with the original

VRP. Indeed the energy used by VRP with load bal-

ance is almost the same as DSR and AODV. Figure

5 depicts only the energy used by the routing proto-

col to deliver each data message. Even in this case it

is possible to notice that the load balance technique

reduces the VRP energy consumption.

Figures 6, 7 and 8 depict the results varying the

maximum speed of the units. Even though ﬁgure 6

shows that DSR and AODV still have a better network

lifetime, the use of load balance on VRP signiﬁcantly

increases the network lifetime. This shows that the

load balance was able to ﬁnd a route with more resid-

ual energy, thus saving the units with less energy from

participating in the routing process.

Figure 3: Network density: network lifetime.

Figure 4: Network density: total energy to deliver a data

message.

Figure 5: Network density: energy used by the routing pro-

tocol to deliver a data message.

Figure 6: Maximum speed: network lifetime.

The energy necessary to deliver a data package

(Figure 7) is higher in VRP due to the overhead

caused by the routing information sent within the data

package and by the need of constant proactive scout

update. In spite of that, the use of load balance over

APPLYING LOAD BALANCE TO REDUCE THE ENERGY CONSUMPTION OF THE VIRTUAL ROUTING

PROTOCOL

the VRP increases the network lifetime specially to

speeds over 4m/s. Note that the energy used exclu-

sively by the routing protocol messages to deliver a

data message (Figure 8) is not inﬂuenced by the use of

load balance. This shows that the load balance tech-

nique was able to select better routes, i.e., routes with

more residual energy.

Figure 7: Maximum speed: total energy to deliver a data

message.

Figure 8: Maximum speed: energy used by the routing pro-

tocol to deliver a data message.

Figure 9: Throughput: network lifetime.

Figures 9, 10 and 11 show the simulations results

while varying the throughput of the network. In Fig-

ure 9, it is possible to notice that the use of load bal-

ance increases the network lifetime of VRP and it is

able to maintain the gain even in high trafﬁc condi-

tions. Besides, the use of load balance made the re-

sults of VRP very similar to the ones of DSR and

VRP, independently of the throughput. Furthermore,

the energy necessary to deliver data package (Fig-

ure 10) is practically the same for VRP with load bal-

ance and AODV and DSR, all these without changing

the energy used exclusively by the routing protocol

messages (Figure 11).

Figure 10: Throughput: total energy to deliver a data mes-

sage.

Figure 11: Throughput: energy used by the routing protocol

to deliver a data message.

5 CONCLUSIONS AND FUTURE

WORK

Routing and reducing the energy consumptionare two

of the most critical challenges of a mobile ad hoc net-

work. Furthermore, both tasks are interdependent,

as turning off the network interface to save energy,

routing protocols are immediately affected. Therefore

routing protocols and energy saving algorithms must

be designed to work within each other.

A technique to save energy without turning off the

network interface of the units is the load balance tech-

nique. The load balance technique is applied to the

routing protocol, which must avoid the use of units

with low residual energy. The use of the load balance

technique might distribute the energy consumption

uniformly over the network, thus maximizing the net-

work lifetime without turning off any network inter-

face. This article applies the concept of load balance

to the Virtual Routing Protocol. The Virtual Routing

Protocol is a hybrid routing protocol for MANETs,

which uses a virtual structure to route packages over

the network.

WINSYS 2009 - International Conference on Wireless Information Networks and Systems

Applying the load balance technique to the Vir-

tual Routing Protocol, it is possible to increase the

network lifetime, without a signiﬁcantly impact on

the delivery ratio. As shown, the energy used ex-

clusively by the routing protocol messages to deliver

a data message is not inﬂuenced by the use of load

balance. This shows that the load balance technique

was able to select better routes, i.e., routes with more

residual energy, to send data messages.

Future work includes the study of the load balance

technique over the Virtual Distance Vector, to verify

if it is possible to achieve the same gain without los-

ing its major advantage, a very small delay to build a

route. It also includes the study of the applicability

of the load balance with multiple paths on the Virtual

Routing Protocol and the Virtual Distance Vector.

REFERENCES

Albini, L., Caruso, A., Chessa, S., and Maestrini, P. (2006).

Reliable routing in wireless ad hoc networks: The vir-

tual routing protocol. Journal of Network and Systems

Management, 14(3):335–358.

Bannack, A. and Albini, L. (2008). Investigating the load

balance of multi-path routing to increase the lifetime

of a manet. Proceedings of the International Con-

ference On Circuis and Systems for Communications

(ICCSC 2008), pages 109–113.

Cho, H. K., Kim, E. S., and Kang, D. W. (2005). A load-

balancing routing considering power conservation in

wireless ad-hoc networks. In Proceedings of the Inter-

national Workshop on Database and Expert Systems

Applications, pages 128–132.

Feeney, L. M. and Nilsson, M. (2001). Investigating the en-

ergy consumption of a wireless network interface in

an ad hoc networking environment. In Proceedings of

IEEE Computer and Communications Societies (IN-

FOCOM 2001), volume 3, pages 1548–1557.

Ganjali, Y. and Keshavarzian, A. (2004). Load balancing in

ad hoc networks: Single-path vs multi-path routing.

In Proceedings of the IEEE Conference on Computer

Communications (INFOCOM 2004), pages 352–355.

Hundewale, N., Jung, S., and Zelikovsky, A. (2007). Energy

efﬁcient node caching and load balancing enhance-

ment of reactive ad hoc routing protocols. Journal

of Universal Computer Science, vol. 13:110–132.

Iqbal, M., Gondal, I., and Dooley, L. (2006). A novel load

balancing technique for proactive energy loss mitiga-

tion in ubiquitous networks. In Proceedings of the

IEEE International Conference on Consumer Com-

munications and Networks (CCNC 2006), volume 1,

pages 157–167.

Johnson, D. and Maltz, D. (1996). Dynamic source routing

in ad hoc wireless networks. In Imielinski and Ko-

rth, editors, Mobile Computing, volume 353. Kluwer

Academic Publishers.

Lee, S. and Gerla, M. (2001). Split multipath routing with

maximally disjoint paths in ad hoc networks. Proceed-

ings of the IEEE International Conference on Commu-

nications (ICC 2001), 10:3201–3205.

Perkins, C. and Royer, E. M. (1999). Ad-hoc on-demand

distance vector (AODV) routing. In Proceedings of

the IEEE Workshop on Mobile Computing Systems

and Applications (WMCSA 1999), pages 90–100.

Robba, A. and Maestrini, P. (2007). Routing in mobile ad-

hoc networks: The virtual distance vector protocol.

In Proceedings of the IEEE International Conference

on Mobile Ad-hoc and Sensor Systems (MASS 2007),

pages 1–9.

Ziane, S. and Mellouk, A. (2005). A swarm intelligent

multi-path routing for multimedia trafﬁc over mobile

ad hoc networks. In Proceedings of the ACM interna-

tional workshop on Quality of service and security in

wireless and mobile networks (Q2SWinet 05), pages

55–62.

APPLYING LOAD BALANCE TO REDUCE THE ENERGY CONSUMPTION OF THE VIRTUAL ROUTING

PROTOCOL