Future Mobile Communications: LTE Optimization and Mobile Network Virtualization

Providing QoS while optimizing the LTE network in a cost efficient manner is very challenging. Thus, radio scheduling is one of the most important functions in mobile broadband networks. The design of a mobile network radio scheduler holds several objectives that need to be satisfied, for example: the scheduler needs to maximize the radio performance by efficiently distributing the limited radio resources, since the operator’s revenue depends on it. In addition, the scheduler has to guarantee the user’s demands in terms of their Quality of Service (QoS). Thus, the design of an effective scheduler is rather a complex task. In this thesis, the author proposes the design of a radio scheduler that is optimized towards QoS guarantees and system performance optimization. The proposed scheduler is called “Optimized Service Aware Scheduler” (OSA). The OSA scheduler is tested and analyzed in several scenarios, and is compared against other well-known schedulers. A novel wireless network virtualization framework is also proposed in this thesis. The framework targets the concepts of wireless virtualization applied within the 3GPP Long Term Evolution (LTE) system. LTE represents one of the new mobile communication systems that is just entering the market. Therefore, LTE was chosen as a case study to demonstrate the proposed wireless virtualization framework. The framework is implemented in the LTE network simulator and analyzed, highlighting the many advantages and potential gain that the virtualization process can achieve. Two potential gain scenarios that can result from using network virtualization in LTE systems are analyzed: Multiplexing gain coming from spectrum sharing, and multi-user diversity gain. Several LTE radio analytical models, based on Continuous TimeMarkov Chains (CTMC) are designed and developed in this thesis. These models target the modeling of three different time domain radio schedulers: MaximumThroughput (MaxT), Blind Equal Throughput (BET), and Optimized Service Aware Scheduler (OSA). The models are used to obtain faster results (i.e., in a very short time period in the order of seconds to minutes), compared to the simulation results that can take considerably longer periods, such as hours or sometimes even days. The model results are also compared against the simulation results, and it is shown that it provides a