Performance and Comparative Analysis of Design Schemes for

Prioritised Data in Multi-hop Wireless Mesh Backbone Networks

Sajid M. Sheikh, Riaan Wolhuter and G. J. van Rooyen

Department of Electrical and Electronic Engineering, University of Stellenbosch,

Private Bag X1, Matieland, 7602, South Africa

Keywords: DiffServ, EDCA, Fairness, IntServ, MAC, QoS, Routing, Smart Grid, Wireless Mesh Networks.

Abstract: The contention based carrier sense multiple access with collision avoidance (CSMA/CA) was originally

designed for single-hop networks. For CSMA/CA to be used in multi-hop distributed networks and to

provide guaranteed data priority, the CSMA/CA needs to be optimised. An application is the smart grid

consisting of different network domains with data of different priority levels. The IEEE802.11e standard

was developed to provide differentiated data services. With the default enhanced distributed channel access

(EDCA) settings for QoS, an unfairness problem exists for different data classes where higher priority data

can starve lower priority data and also where bandwidth is allocated unfairly. In this paper, we carry out an

investigation of six design schemes for wireless backbone networks for QoS provisioning of different data

priority classes. The design schemes are based on the concept of low-cost design for suitability in rural areas

where cost plays a major role. The simulation results were obtained using OMNET++ and the INET

framework. The performance metrics used for the analysis were end-to-end latency, packet loss percentage

and Jain’s fairness index. Simulation results show that hybrid network designs using distributed

coordination function (DCF) and EDCA can improve QoS in terms of reliability and fairness.

1 INTRODUCTION

Wireless Mesh Networks (WMNs) have gained

increasing popularity and use in recent years. This

comes due to the characteristics of WMNs that

include self organisation, auto configuration and low

cost to extend network coverage. With WMNs,

many challenges are also experienced such as

network capacity analysis, QoS routing, link-layer

resource allocation, network security, and seamless

roaming (Jiang et al., 2006). Much research has been

done and published in various areas of WMNs

which includes routing metrics, optimum routing,

security, scheduling, cross layer designs and

physical layer techniques. The capacity of WMNs is

affected by many factors such as network

architecture, network topology, traffic patterns,

network node density, number of channels used for

each node, transmission power level and node

mobility (Akyildiz et al., 2005).

Currently there are two main categories of MAC

scheduling, namely contention based and contention

free strategies. Carrier sense multiple access with

collision avoidance (CSMA/CA) is a popular

contention based scheme deployed in wireless local

area networks (WLANs) and ad-hoc networks. The

original CSMA/CA cannot perform well in wireless

multi-hop environments and offers poor throughput

performance and unfairness problems (Jiang et al.,

2006). Although IEE802.11 and IEEE 802.11e work

well in single hop networks, they present significant

challenges when used in ad-hoc networks such as

collision problems, throughput degradation and the

collision window being increased significantly with

an increase in collision and hence increasing the

end-to-end delay (Yeh, 2004).

The distributed coordination function (DCF) in

the IEEE 802.11 standard does not provide data

priority. The IEEE 802.11e standard was developed

to provide service differentiation using the enhanced

distributed channel access (EDCA). For the different

access categories (AC) or data classes, different

arbitration interframe spacing (AIFS), different

minimum and maximum contention window (CW)

sizes parameters are used in the backoff procedure

for service differentiation. High priority traffic gets

assigned smaller AIFS and CW values compared to

lower priority data classes. This gives the higher

priority AC a higher chance to access the channel

first compared to the lower AC. EDCA can only

Sheikh S., Wolhuter R. and Rooyen G..

Performance and Comparative Analysis of Design Schemes for Prioritised Data in Multi-hop Wireless Mesh Backbone Networks.

DOI: 10.5220/0005567300130023

In Proceedings of the 12th International Conference on Wireless Information Networks and Systems (WINSYS-2015), pages 13-23

ISBN: 978-989-758-119-9

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

provide increased statistical chances rather than

guaranteed prioritised access to higher priority

traffic (Jiang et al., 2006).

For companies to setup wireless telemetry

networks in rural areas such as for micro-grids or to

extend the smart grid, particularly in Africa, the rate

of return of their investment plays a major role as

the population size is smaller and the standard of

living is lower compared to urban areas with a high

percentage of low income people (Sargunarangan,

2011)(Monitor, 2012). Therefore, in rural areas,

WMNs may be more feasible as a more cost

effective solution compared to other solutions such

as fiber optic, cellular networks, WiMax or VSATs.

A typical use case for WMNs viewed in this paper is

the smart grid. The smart grid is comprised of many

network domains that have to be interconnected to

provide end-to-end services. Data in these different

rural smart grid domains need to be given different

priorities depending on the application domain, such

as smart meter data, data from management or

control domains or monitoring domains (Jeon,

2011).

We investigate six design schemes for the

wireless backbone mesh networks based on the

objective of determining high performance, low cost

design implementations. The idea of investigating

different rules assigned to edge and core routers has

been taken from wired networks that provide

differentiated services using Integrated Services

(InterServ) and Differentiated Services (DiffServ).

The edge routers perform most of the complex

operations and the core routers perform simple

operations. In our investigations, schemes 1, 2 and 3

are based on single radio devices using a single

channel in the complete network. In schemes 4, 5

and 6, the core devices use single radios, while the

edge devices use two radios with two channels. Two

radios are only used in a few devices (edge devices)

to keep cost of implementation low. Hardware that

operate in the unlicensed Industrial, Scientific and

Medical (ISM) band can also provide lower cost as

compared to the use of licensed spectrum. In scheme

1, both the edge and core routers are configured with

the default IEEE802.11e EDCA. In scheme 2, the

edge routers are configured with the default

IEEE802.11e EDCA and the core routers are

configured with DCF CSMA/CA. In scheme 3, the

edge routers are configured with DCF and the core

routers are configured with EDCA. Scheme 4 is

identical with scheme 1, scheme 5 is identical to

scheme 2, and scheme 6 is identical to scheme 3

except that the edge routers use two radios and 2

channels in these designs.

The rest of this paper is organized as follows. In

section 2, a brief background on the smart grid

requirements is presented. In section 3, the

objectives of this research are presented. Section 4

presents a brief overview of some current priority

provisioning techniques. Related work is presented

in section 5. Section 6 presents an overview of the

proposed design schemes. Section 7 presents the

simulation experimental setup details. The

performance results are presented in section 8 for the

proposed design schemes.

2 BACKGROUND

In developed countries, a high percentage of people

are connected to the internet, while in developing

countries in Africa, the case is different. A

significant percentage of people in developing

countries in Africa, particularly in rural areas, are

living without internet. Most of the internet

subscribers in developing countries in Sub-Saharan

Africa are in urban areas. This leaves rural areas in

Africa, particularly Sub-Saharan Africa, mostly

without any internet connectivity (Johnson, 2013).

This creates a digital divide. The internet bandwidth

in rural areas is also very limited due to cost (Argaez

2014) (Johnson et al., 2012). The challenges faced

by rural communities include the lack of

communication infrastructure due to cost to provide

this infrastructure (Johnson et al., 2012) (ITU,

2014).

In most cases, a rural village in Africa is up to a

thousand kilometres away from urban areas and

villages are also widely scattered and separated

(Johnson 2013). The cost of covering this distance to

reach scattered rural villages is very high. As a

result, many rural deployments rely on expensive

satellite links (usually VSATs or cellular networks)

to provide internet access (i Direct n.d.) (Hammond

and Paul, 2006).

In rural areas, lower cost and cost effective

wireless communication based on WMNs may be

more feasible. The settlements in these villages are

usually scattered and found in clusters. The

backbone network can be extended and inter-

connected in a wireless mesh method to service

these clusters or connect these local power

generation sources. Wireless backhaul mesh

networks reduce deployment cost and extend

network coverage. The existence of multiple routes

between source and destination nodes ensures high

network availability when node or link failures occur

or when channel conditions are poor (Madihian,

2007).

WINSYS2015-InternationalConferenceonWirelessInformationNetworksandSystems

Table 1: Smart grid communication requirements.

Priority

Category

End-to-end

Latency

Applications Reliability

HIGH < 500ms Emergency Response, Fault

Detection, SCADA, Operations

Data

99-99.9%

MEDIUM 500ms - 2s Automated Demand Response

(ADR), Direct Load Control,

Transformer Monitoring, Outage

Management

99-99.9%

LOW 2s - 5s Advanced Metering Infrastructure

(AMI), Real Time Pricing, Voltage

and Current Monitoring, Remote

connect/Disconnect

99-99.9%

For this study, the requirements of a smart grid

communication network are considered. The data

services have been grouped into three priority levels

i.e. high, medium and low. For WMNs to be used in

the smart grid, it will be expected to provide the

required QoS as summarised in table 1. The network

must be very reliable and provide end-to-end latency

in communication within the tolerated ranges. These

requirements are the same in both urban and rural

networks. The advanced smart metering

infrastructure can tolerate more delay than network

data from fault detection networks. Detailed smart

grid performance requirements in terms of latency

and reliability for different smart grid applications

are also stated in (Gungor, 2011) and (Jeon, 2011).

3 OBJECTIVES OF THE STUDY

Many networks carry data of different priority data,

require the network to be very reliable and provide

end-to-end latency within the tolerated ranges. In

this paper, WMNs are investigated to provide low

cost backbone connectivity for networks carrying

data of different priority such as for the smart grid in

rural areas as highlighted in section 2.

The objectives of this study are:

• To conduct performance measures for edge

and core routers in different EDCA network

design schemes to provide QoS service level

differentiation.

• To improve the overall network reliability in

a wireless multi-hop mesh network through

hybrid network designs.

• To investigate how CSMA/CA can be

configured to give optimum performance in

multi-hop wireless mesh backbones.

4 PRIORITY PROVISIONING

TECHNIQUES

4.1 Integrated and Differentiated

Services

In wired networks, QoS provisioning is carried out

using Integrated Services (InterServ) and

Differentiated Services (DiffServ). The edge routers

perform most of the complex operations and the core

routers perform simple operations.

IntServ provides services on a per flow basis.

IntServ has three main traffic classes namely, best

effort service, controlled load service and guaranteed

service. The best effort services are characterized by

an absence of a QoS specification and the network

delivers the best quality possible. In the guaranteed

services classes, users are provided with an assured

amount of bandwidth and end-to-end delay. In the

controlled load services class, users get serviced as

close as possible to the one received by a best-effort

service in a lightly loaded network (Mahadevan and

Sivalingam, 1999). With the IntServ, QoS support

mechanisms at the network elements can be

provided by various packet classifying and

scheduling mechanisms such as Class Based

Queuing (CBQ) and Weighted Fair Queuing (WFQ).

The signalling of the flow requirements is done

using the Reservation Protocol (RSVP). The RSVP

protocol carries the QoS parameters from the sender

to the receiver to make resource reservations along

the path (Mahadevan and Sivalingam, 1999).

For DiffServ, flows are aggregated into classes

and are treated according to their class, while

IntServ provides per-flow guarantees. Diffserv does

not need to book resources in advance as compared

to IntServ. DiffServ performs mapping multiple

flows into a few service levels. The 8-bit TOS (Type

of Service) field in the IP header is included to

support packet classification. The TOS byte is

divided into 6 bit Differentiated Services Code Point

(DSCP) field and a 2-bit unused field (Mahadevan

and Sivalingam, 1999). The edge router operates in a

wired system using DiffServ included in packet

classification, packet marking and traffic

conditioning. The core router functions using

DiffServ include packet forwarding based on the

per-hop behavior (PHB) that is associated with the

packet class. DiffServ provides QoS services by

differentiating between service classes. Every class

gets a different Behaviour Aggregate (BA). A BA is

a collection of packets with the same DSCP crossing

a router node in a particular direction. Packets are

PerformanceandComparativeAnalysisofDesignSchemesforPrioritisedDatainMulti-hopWirelessMeshBackbone

Networks

forwarded according to the Per-Hop-Behaviour

(PHB) associated with the DSCP (Bos, 2007). The

edge routers in the network perform the complicated

functions such as traffic classification and

conditioning, and the core network is kept simple

(without per-flow information), which makes

DiffServ scalable (Jiang et al., 2006). Diffserv

provide specific treatment known as per-hop

treatment depending on the class of the packet.

4.2 Enhanced Distributed Channel

Access (EDCA)

In wireless networks, QoS is provided using EDCA

for contention based CSMA/CA. The DCF operates

on a listen before talk principle known as

CSMA/CA. In the DCF mode, if a node has data to

transmit, it first senses the medium before

transmission. If the medium is sensed to be idle for a

time period known as DCF interframe space (DIFS),

the station then performs a backoff procedure where

a slotted backoff time is generated randomly from a

contention window (CW). After this period, if the

medium is found idle, transmission takes place.

If the medium is sensed to be busy, the station

then waits for the channel to become idle for the

DIFS period and then the backoff procedure is

started again. At the first transmission attempt, the

CW is set to the minimum value, CW

min

. For any

unsuccessful transmission, this value is doubled.

When the contention window reaches its maximum

size of 1023, it stays constant until it can be reset to

min

after the successful transmission.

Many networking applications require

differentiated services. This can be done by giving

higher priority data preferential access to the

medium. The IEEE 802.11e standard has been

developed to provide differentiated services for QoS

provisioning. It specifies the use of EDCA and

hybrid coordination function (HCF) (Kaveh

Pahlavan, 2002). EDCA is an extension of DCF and

introduces the concept of access category (AC) for

data types. Data is mapped at the MAC layer into the

corresponding AC. The four access categories are

background (BK), best-effort (BE), video (VI) and

voice (VO). EDCA introduces a new interframe

spacing called Arbitration IFS (AIFS). For each of

the ACs, the corresponding CW values are shown in

table 2.

Figure 1 illustrates the different AC’s, AIFS and

parallel backoff entities in EDCA in a timing

diagram. AC[0] has the shortest AIFS period and

back off range, compared to the lower priority data.

Figure 2 shows the implementation scheduling

Table 2: Parameters of EDCA assigned to each AC

category.

AC AC Type Traffic Type AIFSN CW

min

max

AC[3] AC_BK Background 7 31 1023

AC[2] AC_BE Best Effort 3 31 1023

AC[1] AC_VI Video 2 15 31

AC[0] AC_VO Voice 2 7 15

structure of EDCA. If any queue has data, data is

scheduled after sensing the medium to be idle for the

AIFS period and CW backoff depending on the

priority class. If data from two ACs finish the AIFS

period and CW back off period, an internal collision

takes place. The internal collision is handled by the

virtual collision handler, where the higher priority

data is allowed to transmit and the lower priority

data has to contend for the medium again behaving

as if a collision on the medium as occurred.

Figure 1: EDCA timing diagram.

Figure 2: Reference EDCA implementation model for

IEEE802.11e.

WINSYS2015-InternationalConferenceonWirelessInformationNetworksandSystems

EDCA is characterised by inherent short-term

unfairness (Jiang et al. 2006). One of the reasons for

this unfairness is that when a node transmits

successfully, it sets its CW to the CW

min

, giving its

remaining packets a better chance to be transmitted

before packets from other nodes with a larger CW

(Jiang et al., 2006).

5 RELATED WORK

The fairness problem in IEEE 802.11e has been

mainly addressed in literature using weighted queue

techniques as in (Farn and Chang, 2005), adjusting

CW values as in (Kuppa and Prakash, 2004), fair

queuing as in (Somani and Zhou, 2003) and

(Abuzanat et al. 2009) or adaptive queuing as in

(Hammouri and Daigle, 2011). Very little research

has been conducted in tackling the unfairness and

performance issues presented in EDCA from a

design aspect.

In (Kang, 2006), a differentiated services (DS)

model using IEEE 802.11e in wireless access

networks is proposed. In their design, the wireless

users are able to send and receive RSVP messages.

The wireless access point (WAP) is configured to

carry out IEEE 802.11e service differentiation, carry

RSVP signals and mark packets for service

differentiation in the core. This scheme was

designed mainly for a hybrid model of a wired and

wireless network to provide an end-to-end QoS

guarantee between mobile users over the wireless

access networks.

The novel contributions of this work are that we

introduce a design scheme that differentiates the

roles of edge and core routers. The core routers are

designed to perform simple tasks such as packet

forwarding based on channel contention detection,

while the edge routers are designed to perform more

complex tasks such as data classification and

statistically scheduling data according to the priority

class.

6 PROPOSED DESIGN SCHEMES

In this research, we propose and investigate six

design schemes for the wireless backbone network.

In the proposed schemes we assume a hierarchical

backbone mesh network structure consisting of edge

and core routers. User clients can connect to the

edge routers, while the core routers connect to the

backbone routers and carry the data in the backbone.

Figure 3 shows the six design concepts used in our

investigation.

The schemes are based on the concept of low cost

design implementation solutions and hence we

investigate designs 1, 2 and 3 for single radio and

single channel for both edge and core devices.

Schemes 4, 5 and 6 are the same designs as schemes

1, 2 and 3, with the addition of an additional radio in

the edge nodes and an additional channel. In

schemes 1, 2 and 3, omni-directional antennas are

used. In schemes 4, 5 and 6, omni-directional

antennas are used, with 1 radio in the edge devices

connecting the user devices and the other antenna

connecting the backbone devices. Non-overlapping

channels are used.

In schemes 1 and 4, EDCA is configured in both

the edge and core routers. In schemes 2 and 5, DCF

is configured in the core routers and EDCA is

configured in the edge routers. In schemes 3 and 6,

EDCA is configured in the core routers and DCF is

configured in the edge routers.

7 SIMULATION ENVIRONMENT

To investigate our design concepts, simulations were

set up in OMNET++ using the INET framework.

OMNET++ is an open source application. The INET

framework offers detailed modelling of radio

propagation, interference estimation, implementation

of various MAC, network layer, and transport and

application layer protocols of wireless network

(Ganlenbein, 2010). Table 3 gives details of the

simulation setup implemented in OMNET++ using

the INET framework. In our designs, we assume no

capture effects, and no hidden terminal or exposed

terminal problems. Simulations were carried out on

different network sizes. The maximum network size

for the backbone mesh used in the simulations was

36, as in real life deployments a network using this

many nodes can cover a comparatively large area in

outdoor applications. The standard IEEE802.11e

model with AC[0] for high priority data, AC[1] for

medium priority data and AC[2] for low priority

data was used. The traffic type was heterogeneous

with different priority levels.

For each of the 6 design schemes, 3 experiments

were carried out on different network topology sizes

(3x3, 4x4 and 5x5). Therefore, a total of 18

experiments were carried to obtain the results for the

performance analysis. The experiments were each

repeated twice to verify the results. The confidence

interval was 95%.

PerformanceandComparativeAnalysisofDesignSchemesforPrioritisedDatainMulti-hopWirelessMeshBackbone

Networks

Figure 3: Proposed design schemes for the wireless backbone mesh network.

Table 3: Simulation Environment.

Network Setup

Simulation Time 300 seconds

Topologies type: Grid Topology

Number of Nodes 16, 25 and 36 for the backbone

Mesh Sizes for Backbone nodes 4x4, 5x5, 6x6

Backbone separation Distance 300m between nodes

Area 2.2km x 2.2km = 4.84km

Propagation Model Free Space Model

Carrier Frequency 2.4GHz

Data rate 54Mbits/s

Application Data UDP Basic Burst Packets

Data Categories

3 categories of Data - Low, Medium

and High Priority

Packet Size 512bytes

Rate of Transmission 100 packets/second

User Data Protocol (UDP) data was used in our

simulations for the three types of priority data. UDP

does not establish connections between the source

and destinations (connectionless) and also there are

no retransmission of lost packets (Xylomenos and

Polyzos, 1999). The use of UDP packets help

determine the reliability of the network through

packet loss measures. On the other hand,

Transmission Control Protocol (TCP) is connection

oriented and also feedback is received for delivered

packets (Xylomenos and Polyzos, 1999).

To test the possibility of nodes dropping packets

and also higher priority data starving lower priority

data, the arrival data rate for all the data classes was

set to 100 packets/sec. In many real life situations,

the end devices are usually randomly distributed

which gain access to the network by connecting to

the backbone devices. In rural settings in Africa, it is

possible to layout backbone grid topologies or

topologies that are close to grid topology due to

large open areas as mentioned briefly in section 2.

Grid topologies provide a high number of mesh links

and hence increase the reliability when some node

connections are lost.

To assess the performance of our proposed

design schemes and carry out the comparative

analysis, end-to-end latency, packet loss (%) and

Jain’s Fairness Index metrics were used:

1. End-to-end latency: This is the average time

taken by a data packet to arrive at the destination

from the source. It includes all the delay experienced

from the source to the destination which includes

WINSYS2015-InternationalConferenceonWirelessInformationNetworksandSystems

route discovery processes, data queuing and packet

transmission. Only the data packets successfully

delivered to the destinations are used in these

calculations (Vardakas et al., 2007).

2. Packet Loss: This is the measure of the

percentage of packets lost from the source to the

destination. This value was measured at the

destination as (Periyasamy, 2014):

PacketLoss





=













∗





(1)

Where 



is the number of packets is transmitted

and 



is the number of packets received.

3. Jain Fairness Index (JFI): A fairness index is

a measure of how fair or unfair the resources are

shared among the competing hosts. Equation 2 is

used to calculate fairness where x

is the normalized

throughput of station i, and n is the number of flows

in the WMN. A JFI of 1 indicates absolute fairness

and a JFI of 0 absolute unfair resource distribution

(Deng and Han, 2009). In our case n=3 as we

investigate the fairness for 3 data classes namely for

high, medium and low priority classes.









,



,



,….,





=



∑

















∑













 (2)

ℎ0 ≤ 







,



,



,….,





≤1

8 RESULTS

The performance of the six schemes were analysed

in terms of packet loss in figures 4 to 9. Figure 4

displays the packet loss for high priority data in

schemes 1, 2 and 3. Scheme 2 experienced the least

packet loss for high priority data in a 4 by 4 network

and 5 by 5 network. Scheme 3 experienced the least

packet loss in a 6 by 6 network for high priority

data. For high priority data in schemes 4, 5 and 6 as

can be seen in figure 5, scheme 5 experienced the

least packet loss. Figure 6 displays the packet loss

for medium priority data in schemes 1, 2 and 3.

Scheme 2 experienced the least packet loss for

medium priority data in all investigated topology

sizes. For medium priority data in schemes 4, 5 and

6 as can be seen in figure 7, scheme 5 experienced

the least packet loss. Figure 8 displays the packet

loss for low priority data in schemes 1, 2 and 3.

Scheme 2 experienced the least packet loss for low

priority data in 4 by 4 and 5 by 5 topologies. Scheme

1 experienced the least packet loss in the 6 by 6

topology. For low priority data in schemes 4, 5 and 6

as can be seen in figure 9, scheme 5 experienced the

least packet loss. Overall, in the single radio and

single channel schemes (schemes 1, 2 and 3),

scheme 2 which uses DCF in the core routers and

EDCA in the edge routers performed the best. In the

two radio and two channel schemes (schemes 4, 5

and 5), scheme 5 which uses DCF in the core routers

and EDCA in the edge routers performed the best in

terms of least packet loss. For both schemes 2 and 5,

DCF is configured in the core routers and EDCA in

the edge routers. DCF is the core routers gives all

packets carried in the core network an equal chance

of medium access and also packets are transmitted in

a first in first out (FIFO) fashion in the core network.

Doing this reduces the collision probability for high

and medium priority data. The performance in terms

of packet loss reduction improves with the addition

of the additional resources in the edge nodes as in

scheme 5. Edge routers in real life networks are

usually subjected to more traffic load and

congestion. The multi-radio and multi-channel

scheme used in scheme 5 helps lower the packet loss

considerably. Packet loss reduces in the hybrid

design of DCF in the core routers and EDCA in the

edge routers as the number of collisions is expected

to have reduced due to a larger CW range in the core

routers. DCF have larger CW ranges and contention

periods compared to the differentiated IEEE802.11e

services differentiation scheme. Higher range values

of CW with larger back off intervals reduce the

collision probability.

Figures 10 to 16 present the end-to-end latency

experienced for the six design schemes. Figure 10

shows the end-to-end latency for high priority data

in schemes 1, 2 and 3. It can be seen that high

priority data in scheme 1 experienced the least end-

to-end latency compared to schemes 2 and 3. Figure

11 shows the end-to-end latency for schemes 4, 5

and 6 for high priority data. Schemes 5 and 6 high

priority data experienced more latency compare to

scheme 4. The single channel, single radio schemes

(1, 2 and 3) experience more delay then the 2

channel and 2 radios in the edge devices (schemes 4,

5 and 6). Figure 12 shows the end-to-end latency in

schemes 1, 2 and 3 for medium priority data.

Schemes 2 and 3 medium priority data experienced a

considerable increase in latency compared to scheme

1. In figure 13 for medium priority data for schemes

4, 5 and 6, an increase in latency can be observed for

schemes 5 and 6 compared to scheme 4. In figure 14

for low priority data for schemes 1, 2 and 3, schemes

2 and 3 also experience an increase in latency

compared to scheme 1. The increase in latency for

schemes 2 and 3 for low priority data is not as much

PerformanceandComparativeAnalysisofDesignSchemesforPrioritisedDatainMulti-hopWirelessMeshBackbone

Networks

as that experienced for high and medium priority

data. With the hybrid design schemes with the DCF,

each priority data class are given a fixed DIFS and

CW backoff interval which results in an increase in

end-to-end latency for high and medium priority

data. It can be observed that for the cases of EDCA

configured in both edge and core routers, the latency

experienced was the least. For the schemes where

DCF was configured in the core routers and EDCA

was configured in the edge routers, a higher end-to-

end latency was experienced then the scheme with

all EDCA configured routers. The scheme where

DCF was configured in the core routers and EDCA

was configured in the edge routers, also experienced

higher end-to-end latency compared to the scheme

where EDCA was configured in the core routers and

DCF in the edge routers.

Figure 4: Packet loss for high priority data for schemes 1,

2 and 3.

Figure 5: Packet loss for high priority data for schemes 4,

5 and 6.

Figure 6: Packet loss for medium priority data for schemes

1, 2 and 3.

Figure 7: Packet loss for medium priority data for schemes

4, 5 and 6.

Figure 8: Packet loss for low priority data for schemes 1, 2

and 3.

Figure 9: Packet loss for low priority data for schemes 4, 5

and 6.

Figure 10: End-to-end latency measured for high priority

data in schemes 1, 2 and 3.

0,0

20,0

40,0

60,0

80,0

4 x 4 5x5 6x6

Packet Loss (%)

Toplogy Size

Scheme 1 Scheme 2 Scheme 3

0,0

10,0

20,0

30,0

40,0

4 x 4 5x5 6x6

Packet Loss (%)

Toplology Size

Scheme 4 Scheme 5 Scheme 6

0,0

20,0

40,0

60,0

80,0

4 x 4 5x5 6x6

Packet Loss (%)

Topology Size

Scheme 1 Scheme 2 Scheme 3

0,0

10,0

20,0

30,0

40,0

4 x 4 5x5 6x6

Packet Loss (%)

Topology Size

Scheme 4 Scheme 5 Scheme 6

0,0

20,0

40,0

60,0

80,0

4 x 4 5x5 6x6

Packet Loss (%)

Topology Size

Scheme 1 Scheme 2 Scheme 3

0,0

10,0

20,0

30,0

4 x 4 5x5 6x6

Packet Loss (%)

Toplogy Size

Scheme 4 Scheme 5 Scheme 6

0,0

10,0

20,0

30,0

40,0

50,0

60,0

70,0

80,0

90,0

4 x 4 5x5 6x6

End-to-End Latency (msec)

Topology Size

Scheme 1

Scheme 2

Scheme 3

WINSYS2015-InternationalConferenceonWirelessInformationNetworksandSystems

Figure 11: End-to-end latency measured for high priority

data in schemes 4, 5 and 6.

Figure 12: End-to-end latency measured for medium

priority data in schemes 1, 2 and 3.

Figure 13: End-to-end latency measured for medium

priority data in schemes 4, 5 and 6.

With the default EDCA, an unfairness problem

exists as mentioned where higher priority data can

starve lower priority data. The Jain’s fairness index

for all the six schemes is shown in figure 16. For the

single radio and single channel schemes (schemes 1,

2 and 3), scheme 2 provided the highest fairness.

Figure 14: End-to-end latency measured for low priority

data in schemes 1, 2 and 3.

Figure 15: End-to-end latency measured for low priority

data in schemes 4, 5 and 6.

Figure 16: Fairness indication of the three schemes using

Jain’s Fairness Index.

The two radios, two channels schemes configured

in the edge routers, all provided fairness of 1. The

network design in scheme 2 is shown to provide

higher fairness. Schemes 2 and 5 give an equal

chance probability to all data priority classes in the

core network which improves fairness in the

network.

0,0

1,0

2,0

3,0

4,0

5,0

6,0

4 x 4 5x5 6x6

End-to-End Latency (msec)

Topology Size

Scheme 4

Scheme 5

Scheme 6

0,0

10,0

20,0

30,0

40,0

50,0

60,0

70,0

80,0

90,0

4 x 4 5x5 6x6

End-to-End Latency (msec)

Toplogy Size

Scheme 1

Scheme 2

Scheme 3

0,0

1,0

2,0

3,0

4,0

5,0

6,0

7,0

8,0

4 x 4 5x5 6x6

End-to-End Latency (msec)

Toplogy Size

Scheme 4

Scheme 5

Scheme 6

0,0

10,0

20,0

30,0

40,0

50,0

60,0

70,0

80,0

90,0

100,0

4 x 4 5x5 6x6

End-to-End Latency (msec)

Toplogy Size

Scheme 1

Scheme 2

Scheme 3

0,0

2,0

4,0

6,0

8,0

10,0

12,0

4 x 4 5x5 6x6

End-to-End Latency (msec)

Topology Size

Scheme 4

Scheme 5

Scheme 6

0,950

0,955

0,960

0,965

0,970

0,975

0,980

0,985

0,990

0,995

1,000

1,005

4 x 4 5x5 6x6

Jain's Fairness Index

Topology Size

SCHEME 1

SCHEME 2

SCHEME 3

SCHEME 4

SCHEME 5

SCHEME 6

PerformanceandComparativeAnalysisofDesignSchemesforPrioritisedDatainMulti-hopWirelessMeshBackbone

Networks

9 CONCLUSIONS

In this paper the use of CSMA/CA in multi-hop

distributed backhaul networks to provide guaranteed

data priority under different design schemes was

investigated. We investigated the performance of six

design schemes for wireless backbone mesh

networks. Different roles were assigned to the edge

and core routers. Schemes 1, 2 and 3 used single

radio and single channel in core and edge routers. In

Scheme 1, all routers performed the same role and

were configured with EDCA. In Scheme 2, the edge

routers performed data classification and were

configured with EDCA. The core routers in Scheme

2 were only configured with the default DCF. In

scheme 3, DCF was configured in the edge routers

and EDCA in the core routers. Schemes 4, 5, and 6

were identical to schemes 1, 2 and 3 with the

addition of another radio in the edge routers and an

additional channel in the network.

The hybrid design scheme where DCF was

configured in the core routers and EDCA in the edge

routers experienced the least packet loss. This was

due to a reduction in the number of collisions as

DCF have larger CW ranges and contention periods

compared to the differentiated IEEE802.11e services

differentiation scheme. Higher range values of CW

with larger backoff intervals reduce the collision

probability. The different data packets carried in the

backbone devices with DCF configured have an

equal chance of gaining access to the medium and

the scheduling of packets operate as a FIFO

scheduling in the backbone devices.

The scheme with all routers configured with

EDCA in both edge and core routers, experienced

the least latency. This is as a result of the service

differentiation with higher priority data waiting less

time to access the medium. The schemes where DCF

was configured in the core routers and EDCA was

configured in the edge routers, experienced higher

delay then the EDCA scheme with all EDCA

routers. For the single radio and single channel

schemes (schemes 1, 2 and 3), scheme 2 provided

the highest fairness.

Networks that require high reliability, but can

tolerate more end-to-end latency, a hybrid design

scheme, where DCF is configured in the core routers

and EDCA is configured in edge routers will be a

good design to use. Rural smart grid networks can

be a potential application for this design scheme.

The fairness problem in IEEE802.11e in

literature has been mainly addressed using weighted

queues, adjusting CW values adaptively and

differentiated services models among others. The

novelty of this work was the improvement of

fairness from a design aspect by assigning different

roles to edge and core devices.

The objectives of the paper have been met where

DCF configured in the core routers and EDCA

configured in the edge nodes provides a hybrid

design scheme that is more reliable with less packet

loss compared to a design with EDCA configured in

all nodes. This hybrid design scheme also provides

more fairness for data of different priority. Hybrid

design schemes reduce collisions and hence result in

improved throughput.

Edge routers are subjected to more traffic load

and congestion in networks. The multi-radio and

channel scheme at the edge routers helps prevent

congestion. Further work would entail developing

fair scheduling schemes.

REFERENCES

Abuzanat, H., Trouillet, B. and Toguyeni, A., 2009. Fair

Queuing Model for EDCA to Optimize QoS in Ad-

Hoc Wireless Network. 2009 Eighth International

Conference on Networks, pp.306–311. Available at:

http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?

arnumber=4976692 [Accessed October 24, 2014].

Akyildiz, I.F., Wang, X. and Wang, W., 2005. Wireless

mesh networks: a survey. Computer Networks, 47(4),

pp.445–487. Available at: http://linkinghub.elsevier.

com/retrieve/pii/S1389128604003457 [Accessed May

2, 2014].

Argaez, E. de, 2014. Internet World Stats. Available at:

http://www.internetworldstats.com/stats1.htm

[Accessed April 11, 2014].

Bos, G., 2007. QoS support using DiffServ. 6th TSConIT.

Deng, J. and Han, Y.S., 2009. Fairness Index Based on

Variational Distance. Global Telecommunications

Conference, pp.1 –6.

Farn, J. and Chang, M., 2005. Proportional Fairness for

QoS Enhancement in IEEE 802.11e WLANS.

International Conference on Local Computer

Networks, (1), pp.4–5.

Ganlenbein, R., 2010. Virtual Mesh: An Emulation

Framework for Wireless Mesh Networks in Omnet++.

University of Bern.

Gungor, P.V.C., 2011. Smart Grid Communications :

Research Challenges and Oppurtunities. Presentation

at Bahcesehir Univeristy, Turkey.

Hammond, A. and Paul, J., 2006. A New Model for Rural

Connectivity. World Resouces Institure, (May).

Hammouri, M.M. and Daigle, J.N., 2011. A Distributed

Scheduling Mechanism to Improve Quality of Service

in IEEE 802. 11 Ad Hoc Networks. IEEE Symposium

on Computers and Communications (ISCC), pp.1–6.

i Direct, 8 Ways to Implementing Backhaul over Satellite

for Mobile Operators.

WINSYS2015-InternationalConferenceonWirelessInformationNetworksandSystems

ITU, 2014. ICTs Go Rural. Available at:

https://www.itu.int/ITU-D/ict_stories/themes/e-

rural.html [Accessed April 11, 2014].

Jeon, Y.-H., 2011. QoS Requirements for the Smart Grid

Communications Systems. International Journal of

Computer Science and Network Security, 11(3),

pp.86–94.

Jiang, H. et al., 2006. Differentiated Services for Wireless

Mesh Backbone. Communications Magazine, IEEE,

44(7), pp.113–119.

Johnson, D.L., 2013. Re-architecting Internet Access and

Wireless Networks for Rural Developing Regions.

PhD Dissertation, (March).

Johnson, D.L. et al., 2012. The Bandwidth Divide :

Obstacles to Efficient Broadband Adoption in Rural

Sub-Saharan Africa The Bandwidth Divide : Obstacles

to Efficient Broadband Adoption in Rural Sub-

Saharan Africa. , 6, pp.2467–2491. Available at:

https://www.smartgrid.gov/federal_initiatives/featured

_initiatives/usda_aims_invest_250_million_rural_sma

rt_grid_deployment.

Kang, M., 2006. Performance Analysis of An Efficient

QoS Scheme over IEEE 802. 11e Based Wireless

Access Networks. International Conference on

Advanced Communication Technology, 2, pp.929–933.

Kaveh Pahlavan, P.K., 2002. Principles of Wireless

Networks,

Kuppa, S. and Prakash, R., 2004. Service differentiation

mechanisms for IEEE 802. 11-based wireless

networks. Wireless Communications and Networking

Conference, 4, pp.796–801.

Madihian, M., 2007. Multi-hop wireless backhaul

networks: a cross-layer design paradigm. IEEE

Journal on Selected Areas in Communications, 25(4),

pp.738–748. Available at: http://ieeexplore.ieee.org/

lpdocs/epic03/wrapper.htm?arnumber=4205056.

Mahadevan, I. and Sivalingam, K.M., 1999. Quality of

Service Architectures for Wireless Networks : IntServ

and DiffServ Models. InternationalSymposium on

Parallel Architectures, Algorithms, and Networks,

pp.420 –425.

Monitor, D.S., 2012. Rural Infrastructure in Africa.

African Monitor, (1).

Periyasamy, P., 2014. Comparative Performance Analysis

of AODV and AODV-MIMC Routing Protocols for

Mobile Ad hoc Networks. International Jounal

Computer Network and Information Security, (May),

pp.54–60.

Sargunarangan, D., 2011. A Low-Cost Efficient Wireless

Architecture for Rural Network Connectivity. ACM

SIGCOMM.

Somani, A.K. and Zhou, J., 2003. Achieving Fairness in

Distributed Scheduling in Wireless Ad-Hoc Networks.

Performance, Computing, and Communications

Conference, 1, pp.95–102.

Vardakas, J.S. et al., 2007. On the End-to-End Delay

Analysis of the IEEE 802. 11 Distributed Coordination

Function. International Conference on Internet

Monitoring and Protection, pp.5–9.

Xylomenos, G. and Polyzos, G.C., 1999. TCP and UDP

Performance over a Wireless LAN. IEEE INFOCOM,

(March), pp.439–446.

Yeh, Ch.-H., 2004. A New Scheme for Effective MAC-

layer DiffServ Supports in Mobile Ad Hoc Networks

and Multihop Wireless LANs. Vehicular Technology

Conference, 4, pp.2149–2155.

PerformanceandComparativeAnalysisofDesignSchemesforPrioritisedDatainMulti-hopWirelessMeshBackbone

Networks