The advent of Internet protocol television (IPTV), Internet gaming and similar bandwidth hungry new generation applications has continued to drive the need for higher bandwidth capacities and by extension the need for service providers to further over-provision network resources (especially in the absence of accurate traffic models). This leads to increased capital expenditure (CAPEX) and reduced RoI. The need for higher bandwidth capacities is most prevalent in the access network, which represents the part of the service provider’s network infrastructure that is closest to the customer. Unfortunately, this part of the service provider’s network has also remained the bottleneck in efforts to deliver bandwidth-intensive new-generation applications and services directly to customers. In the wired access network, GEPON is a promising technology for relieving this bottleneck while its counterpart in the wireless access network is WiMAX. Interestingly, these two promising technologies are quite similar and so synergistically converging them makes very good business sense for relieving the prevailing access network bottleneck issue. Consequently, we propose a converged quadruplet-service (video, voice, data and mobility) capable network that emerges from the convergence of GEPON and WiMAX and we address the technical challenges involved in such a convergence. Such a converged network takes full advantage of the strengths and weaknesses of each of the two promising technologies; a solution that is currently generating a lot of interest in both the academia and in the telecommunications industry.
Meanwhile, in the telecommunications industry, service providers are now moving from a network-centric-based approach to a customer-centric-based approach in terms of service provisioning, which is in line with recognizing the customer as King. Providing customer-centric services involves providing services that meet the quality of service (QoS) expectations of applications that customers use and/or the quality of experience (QoE) expectations of the customers themselves. Further, offering guaranteed or expected QoS or QoE is closely tied to meeting service level agreements (SLAs). SLAs on their part are based on various QoS parameters such as delay, packet loss rate, throughput, etc. To provide profitable customer-centric services therefore, service providers need to use accurate models that will enable them to properly understand the true behavior of QoS parameters in their network and therefore, enable them to properly optimize their network infrastructure for meeting the QoS expectations of their customers. The use of such accurate models will enable them to make higher return on investment (RoI) by limiting over-provisioning of network resources which is what applies in the absence of accurate models. In this regard, reliable research and Internet measurements have showed that in modern networks, actual Ethernet and wireless data traffic, in essence Internet-like traffic, exhibit self-similar and long range dependent (LRD) characteristics. These are attributes that simple Poisson models which have been relied on for several years fail to capture; thus leading to wrong estimates of delay, packet loss rate and other QoS parameters. Accordingly, this results in poor network planning and over-provisioning when service providers rely on such Poisson-based models. To overcome the limitations of current state-of-the-art work, therefore, our analytical modeling of the proposed converged architecture is based on self-similar and long range dependent traffic conditions.
Unlike existing work (which only considers outbound data traffic), our analysis takes into account both inbound and outbound data traffic. It is utmost important to give attention to inbound traffic because most of the inbound traffic is p2p and we can witness the growing popularity and significance of p2p traffic in real networks. Further, unlike existing work that focus on analyzing the queuing system of a converged network based on the simple first-come-first-serve (FCFS) scheduling policy, this dissertation presents for the first time (to the best of our knowledge), the queuing analysis of the converged network under Internet-like traffic conditions using common scheduling disciplines that are likely to be used in future and real networks ? priority queuing (PQ), low latency queuing (LLQ) and custom queuing (CQ). Our analytical models are based on a G/M/1 queuing system that takes into account multiple classes of Internet-like input traffic. We derive exact QoS parameters, such as the expected queue length, expected waiting time in queue (queuing delay), end-to-end delay and the packet loss rate; all per QoS traffic class for the PQ, LLQ and CQ scheduling logics. In addition, we also develop the finite queue Markov chain for each of these queuing schemes. The derived expressions are evaluated through comprehensive numerical analysis and further validated through extensive simulation experiments; which have been conducted to investigate and understand the behavior of self-similar traffic in the converged network and particularly, to see how QoS parameters are affected. This work helps the understanding of the converged network and leads to accurate and exact parameters for predicting QoS behavior in the converged network, hence facilitating the provisioning of tightly bound QoS parameters to end-users. This work can be used as a guide for the efficient allocation and optimization of network resources such as queue size and bandwidth for individual traffic classes for the purpose of guaranteeing the QoS/QoE required by different applications/subscribers.
On the other hand, there has been a rapid growth in the number of mobile subscribers and devices, which has resulted in an increased demand for “anywhere, anytime, and any way” high-speed Internet access. Satisfying this demand requires realizing mobile and ubiquitous computing environments. In this regard, much effort is being put into realizing all-IP mobile networks. All-IP mobile networks will tightly combine the Internet and telecommunication networks together. They are networks in which Internet protocol (IP) is employed from mobile subscriber stations to the access points that connect the wireless networks to the Internet. One of the most critical and challenging issues in all-IP mobile networks is IP mobility management. IP mobility management involves enabling an all-IP mobile network to locate a mobile subscriber’s IP-level point of attachment (i.e., location management) for delivering data packets to it from its correspondent nodes as well as maintaining reachability with the mobile subscriber as it continues to move and change its point of attachment (i.e. handover management). In this regard, many IP mobility management models have been proposed but the time for handover due to the standard operational procedures of these models remains unacceptable to real-time traffic. Moreover, reduction in the time for handover is also highly beneficial to non real-time throughput sensitive applications as well. In this Dissertation, therefore, we also present and analyze a simple and easily implementable proactive QoS-Aware proxy mobile IPv6 model which relies on a rich set of informational resources including the on-the-fly QoS requirements and service level agreements of mobile subscriber devices to make efficient proactive handover decisions. Additionally, we present quantitative and comparative analysis between our model and representative models in a converged optical-wireless access network environment revealing the enormous handover time improvement our model brings over existing models.