Performance Evaluation of Routing Metrics for Wireless Mesh Networks

The objective of this study is to evaluate the performance of existing routing metrics for wireless mesh networks with the aim of designing an optimal one. The objective of this work was achieved by evaluating the performance of existing routing metrics using NS2 simulation and recommending design criteria for designing optimal routing metrics for WMNs.

INTRODUCTION

P REAMBLE
B ACKGROUND

Reactive Routing Protocols

S TATEMENT OF THE P ROBLEM
R ATIONALE OF THE S TUDY
R ESEARCH G OAL AND O BJECTIVES

Research Goal
Research Objectives

O VERVIEW OF R ESEARCH M ETHODOLOGY

Primary Research Method: Simulation
Secondary Research Method: Case-Study

O RGANIZATION OF THE D ISSERTATION

Case study was used to evaluate the performance of the four route metrics simulated. The performance of each routing metric was compared to the performance of the other routing metrics that were also simulated.

Figure 1.1: Types of Multi-hop Wireless Networks

LITERATURE REVIEW

I NTRODUCTION

Most routing metrics comparisons compared routing metrics in an inconsistent manner. Section 2.2 discusses existing comparisons of existing routing metrics, while Section 2.3 concludes the chapter.

Figure 2.1: Relationship between a routing protocol and a routing metric

E XISTING C OMPARISONS OF R OUTING M ETRICS

Designing Routing Metrics for Mesh Networks [Yang et al, 2006]
A Comprehensive Comparison of Routing Metrics for Wireless Mesh Networks [Liu et al, 2008]

The same performance metrics (maximum channel utilization, network throughput, end-to-end packet delay) were used for all routing metrics. Our work adopted the approach used by [Liu et al, 2008] to first cluster the routing metrics.

Table 2.1: Existing routing metrics reviews

REVIEW OF EXISTING ROUTING METRICS

I NTRODUCTION

EETT first checks the busy degree of the link channel before making a route selection. Up to twenty routing metrics were discussed in this chapter, of which four routing metrics were selected for evaluation.

E XISTING R OUTING M ETRICS

Custom Specification for Routing Metrics

S PECIFICATION OF R OUTING M ETRICS

The improved expected transmission time routing metric has been proposed to address the shortcomings of the ETT routing metric. The iETT routing metric is designed to account for (1) the difference in link loss rates within a single route and (2) the overheads of the MAC layer when calculating an expected packet transmission time (rather than simply using packet or bandwidth) .

Table 3.2: Classification of routing metrics [Nxumalo et. al., 2009]

S UMMARY

I NTRODUCTION

EETT takes into account other things that the other routing metrics don't take into account, such as channel distribution on long paths.

H OP C OUNT

E XPECTED T RANSMISSION C OUNT

The flowchart shows the steps that the ETX routing metric follows when it selects a path to use for sending packets.

P ER - HOP R OUND T RIP T IME

E XCLUSIVE E XPECTED T RANSMISSION T IME

The flow chart clearly shows how exclusive expected transmission time routing metrics are involved in selecting an optimal path for sending packets.

S UMMARY

PERFORMANCE EVALUATION OF SELECTED ROUTING METRICS

I NTRODUCTION

The assessment was based on the framework that formed the basis for grouping the twenty routing statistics into four categories (see Chapter Three). One and only one of the routing metrics in each of the four categories (see Chapter Four) was selected for simulation experiments as described in Chapter Three. The selected routing statistics are now ready in the form of pseudocode and flowcharts for simulation experiments in this chapter.

It is better to compare routing statistics using a preferred routing protocol in the literature. This section briefly describes the type of simulation tool and the version of the simulation tool. Traffic volume increased as the number of nodes in the network increased.

S IMULATION E NVIRONMENT

E VALUATION P ARAMETERS

Average delay
Throughput
Packet loss ratio
Delay jitter
Packet drop ratio

We represent the total simulation time bysim_time, while recvd_num is the total number of packets received. The packet loss ratio is the percentage of packets sent that never reached their intended destination. We represent the total number of packets sent bysent_num, while the total number of packets received is represented by recvd_num.

Packet drop ratio is the percentage of packets successfully received by the nodes' intended destination, but dropped for other reasons. Packet loss rate differs from packet loss rate in that packet loss rate looks at packets that have been intentionally dropped due to some reasons, while packet loss rate looks at packets that never reached the intended destination node.

E XPERIMENTAL S ETUP

Routing Protocol
Packet Buffer Model
Physical and Data Link Layer

The formula for calculating the packet drop ratio is shown in Equation 5.6: drop_ is the number of all packets dropped, while the total number of packets received is represented by recvd_num. throughput, delay, packet loss ratio, delay jitter, packet drop ratio) the same for each simulation. AODV was chosen because it was used with most of the routing metrics to be simulated; AODV was appropriate as the focus of this study was on the routing metrics rather than the routing protocol. This routing protocol also provides simplicity and the fact that the two routing metrics (hop count and expected transmission count) evaluated in this work have already been used by this routing protocol.

Since the routing protocol uses Expected Transmission Count, which consumes time, it would not take much time and effort to change it to match the other two (Per-hop Round Trip Time and Exclusive Expected Transmission Time) routing metrics, which also consume time and are descendants of the ETX routing metric. An omnidirectional antenna can send data in all directions with the same transmission strength. One of the advantages of using an omnidirectional antenna is that a node can broadcast packets in any direction with the same strength; therefore, there is no one direction that is better than other directions in the network.

S IMULATION E XPERIMENTS

Experiment 1: The effect of network size on delay
Experiment 2: The effect of network size on delay jitter
Experiment 3: The effect of network size on packet loss ratio
Experiment 4: Effect of network size on packet drop ratio
Experiment 5: Effect of time on throughput
Summary of Results

ETX's low packet loss ratio can also be attributed to the fact that there is less packet loss (see Figure 5-3), since ETX first measures the loss ratio of the link. The aim of this experiment was to determine the throughput of the network over time. An ideal routing metric for WMNs has the potential to increase network throughput;

The high throughput of sending packets based on the hop count can be attributed to the fact that it does not consider the quality of the link when selecting the path along which to send the packet. ETX's best performance can be attributed to the fact that it is designed to improve network throughput. Poor performance due to the hop count can be attributed to the fact that the quality of the link is not taken into account.

R ECOMMENDATION OF D ESIGN C RITERIA

Weight path-awareness
Efficient weight path algorithm design
Quality of Service-awareness
Network scalability

An optimal routing metric for wireless mesh networks should be able to increase the weight of a. An ideal routing metric for WMNs should be able to capture the packet loss ratio to ensure good performance for the minimum weight path. This feature of designing a wireless network routing metric was also recommended by [Yang et al, 2006].

A routing metric like hop count does not take into account the quality of the link through which it intends to send packets. Hop count routing metric does not consider the quality of the link when choosing a route to use for sending packets. An optimal routing metric for wireless mesh network must take into account the scalability of the network.

S UMMARY

C ONCLUSION

The first objective was fulfilled by reviewing two existing studies comparing routing metrics for wireless mesh networks and reviewing twenty existing routing metrics. The routing measurements were later grouped into four groups; The routing metrics in each group were then compared to each other to find a single routing metric to represent the group during simulation and evaluation. The four selected routing metrics were then implemented and evaluated in NS2, fulfilling objective three.

As a result of the evaluation, we discovered that the hop count performed worse than all the routing metrics compared in this study. From the above, we concluded that ETX outperformed all other routing metrics that were simulated, but based on the evaluation, it cannot be concluded that it is the best routing metric for WMN. Research questions two and three are also answered by recommending design criteria for optimal routing metrics for wireless complex networks.

L IMITATIONS AND F UTURE W ORK

The design criteria that informed this study can help guide any scholar who wants to design a new optimal routing metric for WMNs from scratch or extend already existing routing metrics. One routing protocol (AODV) was used in all the simulation experiments performed in this study, another routing protocol can be used to perform the same experiments as well. Hybrid wireless mesh protocol is a routing protocol for WMNs, it should be considered as the second routing protocol to use in this study.

Only one routing metric from each group could be simulated due to time constraints, as the code for the other three routing metrics (RTT, ETX and EETT) had to be hard-coded before they could be used, which took a lot of time to achieve. Selecting more than one route metric from each group would take even more time. As future improvement of this study, the simulation of two route metrics of each group should be considered.

C ONTRIBUTION TO K NOWLEDGE

Proceedings of the First International Conference on Broadband Networks (BROADNETS’04), Washington, DC, USA, October 2004, p. 344-354, Washington, DC, USA: IEEE Computer Society. Proceedings of the 2nd Annual International Conference Workshop on Wireless Internet, Boston, MA, United States, 02-05 August 2006, New York, NY, USA: ACM. In Proceedings of the 10th Annual International Conference on Mobile Computing and Networking, Philadelphia, USA, September 2004, p.

I Proceedings af 2004-konferencen om applikationer, teknologier, arkitekturer og protokoller til computerkommunikation (SZGCOMM), Portland, Oregon, USA, august 2004, s. New York, NY, USA: ACM. Proceedings of the IEEE International Conference on Networking, Sensing and Control, ICNSC 2008, Hainan, Kina, 6.-8. april 2008, s. I Proceedings of the 21st International Conference on Advanced Information Networking and Applications Workshops (AINAW Niagara Falls, Ontario, Canada) , 21-23 maj 2007, s.

In Proceedings of the Second IEEE Workshop on Mobile Computer Systems and Applications, Washington, DC, USA, 1999, pp. Insert that next hop after that RREQ source by that precursor of that RREQ destination rt->pc_insert(rt0->rt_nexthop); // next hop after RREQ source.