Chapter 2 introduces the domain of this study by providing a critical review of related literature. A comprehensive discussion on sensor node and WSN de- ployment architectures providing details of WSN resources and communication channels is presented. Energy conservation mechanisms available for WSNs are analysed. Existing performance and availability models for WSNs are investi- gated. This chapter ends by an introduction of the system under study.
Modelling approaches and solution techniques are critically analysed and com- pared in Chapter 3. Existing performance and availability modelling techniques for communication systems are investigated and critically analysed. In addi- tion, possible solution methods for two-dimensional state spaces are compared. A detailed explanation of System of Linear Equations and Spectral Expansion methods is given.
In chapter 4, an analytical model for a clustered WSN with unbounded queue capacity CH is presented. A queue model for packet arrival distribution at the CH is initially developed. An integrated performance and availability model with inputs taken from arriving data packet is then developed to mimic the system behaviour. The model is subsequently solved using Spectral Expansion solution technique and Poisson approximation solution approach. The results are further validated using results obtained from an event-driven simulation program. In real-life situation, queuing systems do not have infinite queues, In WSN appli-
cations for example, the available memory space is very small hence the amount of data that can be stored is restricted. In chapter 5, bounded queue capac- ities are introduced and incorporated in the model of chapter 4. The model is then used to study performance and availability measures, including Mean Queue Length (MQL), throughput, queuing delay, blocking probability and packet loss. Results are obtained using Spectral Expansion Solution technique and validated using an event-driven simulation program. The obtained results are critically analysed and compared with results attained from independent performance and availability studies.
Sleep scheduling is widely used in WSNs to conserve the limited node energy. While this may save energy significantly, it may also hamper system perfor- mance and increase energy consumption during heavy traffic load when some sleep scheduling schemes are used. In chapter 6, models that employ sleep op- erations are considered. First operation dynamics for sleep schedules in WSNs are critically analysed, and results used to model the system behaviour. The de- veloped models are solved using two analytical approaches: Spectral Expansion and System of linear equations, which are further validated using a dedicated simulation program. Using the developed models, the effects of sleep scheduling on system performance are investigated in terms of MQL, throughput, response time, Failure probability and blocking probability and probability of operating in the various states. Finally, the obtained results are critically analysed and evaluated for purposes of improving system performance.
Like other wireless networks, WSNs are subject to link failures that may arise from environmental, interference and transceiver hardware malfunctions. These eventually degrade system performance hence require consideration. In order to make the models more realistic, chapter 7 presents a complete analytical model for a clustered WSN system with unreliable links in addition to node failures. The model considers a finite queue CH that also conserves energy by entering sleep mode. Both failures are considered repairable and operation restored once repair is complete. Similar to the previous cases, the model is solved using spectral expansion technique and the system of linear simultaneous equations and the results further validated using simulation results. The model is then used to
evaluate system performance in terms of MQL, throughput and response time. Furthermore, channel and node failure rates are varied in order to establish the worst case scenarios that can still allow the system to attain desired performability measures. Additionally, effects of limited queue capacity on system performance are evaluated to establish the blocking probability. Finally, the results obtained from the two solution approached are presented comparatively and validated using simulation results.
Chapter 8 presents a model for energy evaluation based on the models presented in chapter 7. More specifically, the composite model presented in chapter 8 is considered for energy evaluation as a case study. The energy model considers that power is consumed in the various operation states. For this study, mean energy consumed in each operation state is employed for computation of the overall CH energy consumption. The obtained results are further analysed to obtain optimum operation range that provides best power saving for the network. Arrival distributions of packets at the CH in WSNs may vary depending on the application environment and pre-configurations. In chapter 9, a methodology for modelling data arrival distribution for WSNs at the intermediary nodes and the CHs is proposed based on experimental results. First, data delivery mod- els for WSNs are critically analysed. Existing approaches for modelling arrival processes, Bernoulli and Poisson, are critically analysed. Inter-arrival times for data packets are generated using Castalia simulator that runs over OMNET++ platform. Kolmogorov-Smirnov Test is then used to identify if the empirical data follows any known probability distribution and numerical results presented and critical analysed.
Finally, chapter 10 presents a summary of the main contributions of the thesis and outlines some of the possible areas for future studies.
In summary, chapter 1 introduces the domain of this research study by highlight- ing the research question, motivation, objectives and scope of study. A list of publications and an outline detailing work done is also given. The research be- gins by a critical review of relevant literature in chapters 2 and 3. Related works, WSN concepts and system under study are presented in chapter 2 while chapter
3 presents a detailed discussion comparing modelling approaches and possible solution techniques. In both chapters relevant terminologies are introduced. Based on the concepts gathered, requirements of system under study and choices of modelling approaches and solution techniques, the models for the proposed system are progressively developed and solved. Chapter 4 introduces the pro- posed performability modelling approach and uses it to analyse performance of a simple infinite system model with CH breakdowns and repairs. This model is gradually developed in the following chapters to make it more realistic to the actual system. Chapter 5 introduces and analyses the effects of bounded queue capacity on system performance. The model is further advanced in chapter 6 by incorporating and analysing effects resulting from sleep operation dynamics on performance. In addition, chapter 7 extends the model by incorporating link fail- ures in order to evaluate how environmental conditions affect the overall system performance. In chapter 8, an energy model is developed and used to evaluate energy consumption of the developed performability model of chapter 7.
As an extended discussion, chapter 9 presents an initial analysis of packet arrival distribution in WSNs by proposing an approach based on Kolmogorov-Smirnov Test Statistic on empirical data sets. Finally, chapter 10 concludes the work done.
Chapter 2
Literature Review
2.1
Introduction
Considering that WSNs are resource constrained, creating an infrastructure that connects the physical world and gather information for Internet of Things (IoT) adds more complexity to the performance WSN systems. This in turn has in- creased the demand for improved performance and dependability of WSNs in the
wake of many application challenges Aboelaze and Aloul [2005], Akyildiz et al.
[2007]. In many application environments high failure rate exhibited in WSNs limit their lifespan hence affecting their performance and availability thereby degrading overall network QoS. Combined performance and availability studies have successfully been used in modelling communication networks [Franken et al., 1994], [Trivedi et al., 2003], [Gemikonakli et al., 2006], [Do and Chakka, 2010],
[Kirsal et al.,2011] over the years. It is possible to extend the same methodology
for modelling WSNs in order to improve their QoS. Such studies are important in alleviating the impact on performance and availability resulting from frequent WSN node and channel failures. In this Chapter, a detailed literature review covering WSN architecture, existing performance and availability studies, and energy conservation schemes is carried out in order to develop accurate and ef- fective analytical models for performance and availability evaluation of WSNs.
The remainder of the chapter is organised as follows: Section 2.2 presents a
discussion on WSN architecture. Section 2.3 provides a detailed discussion of
energy conservation in WSNs. In section 2.4, QoS and WSN performance are
presented followed by a detailed discussion of the system under study in section
2.5. Finally, the chapter is summarised in section2.6.