Performance evaluation of hierarchical multiservice wireless networks
Student thesis: Doctoral Thesis
Related Research Unit(s)
Wireless cellular networks are evolving to a hierarchical architecture, where small cells (e.g., picocell, femtocell) are covered by macrocells. As capacity in small cells is generally less than that in macrocells, redirecting traffic over owed from lower tier small cells to higher tier macrocells is a promising strategy to improve Quality of Service (QoS) of small cell users. However, QoS requirements of different types of services can be quite different. Generally, traffic generated by end-user services is categorized into delay-sensitive real-time (RT) traffic and delay-insensitive non-real-time (NRT) traffic. To meet QoS requirements of different services and efficiently utilize network resources, typically, RT traffic is given strict priority and NRT traffic shares the remaining capacity. To enable efficient network planning and dimensioning, computational efficient performance evaluation methods are required. However, for the above mentioned hierarchical multiservice wireless networks, performance evaluation is challenging due to the dependency of the NRT traffic performance on the RT traffic and the difficulty in modeling and analysis of the over ow traffic. In this thesis, we study these challenges and provide accurate and computationally efficient performance evaluations for hierarchical wireless networks supporting both RT and NRT traffic. Specifically, for a single cell supporting RT and NRT traffic, we propose two computationally efficient approximation methods called Generalized Quasi-Stationary Approximation (GQSA) and Generalized Fluid Approximation (GFA) for performance evaluation of the NRT traffic in different scenarios. They extend the application range of the previous approximations and provide tradeoffs between computational complexity and accuracy. Meanwhile, we investigate the effect of the NRT call size distributions on the NRT performance and provide insensitive performance bounds and approximations for general NRT call size distributions. This extends the practicality of our analysis. Finally, for the hierarchical network, we propose a Multiservice Equivalent Random Method (MERM) method for fast and accurate modeling and analysis of over ow traffic. Combining MERM with GQSA/GFA, we provide an efficient platform for performance evaluation of hierarchical multiservice wireless networks.
- Wireless communication systems, Evaluation