Mobility Modeling And Resource Management For PCS Networks

The name personal communication services (PCS) is often used as an umbrella term that includes various wireless access and personal mobility services with the ultimate goal of enabling communication, through a small terminal, with a person at any time, at any place, and in any form. Now with the commercial operations of 3G (WCDMA, cdma2000 1xEV-DO, and TD-SCDMA), 3.5G (HSDPA), and 3.75G (HSUPA) communication networks around the world, current PCS networks are ready for fulfilling the grand objective of wireless communications. Unlike ordinary static networks (such as the public telephone network), PCS networks allow the dynamic relocation of mobile terminals. The network access points of mobile terminals change as the users ramble in the coverage of the network. Personal communication implies both terminal mobility and personal mobility, which introduce a wide variety of challenges in teletraffic engineering. This thesis concentrates on three aspects of teletraffic engineering in PCS networks, ranging from call admission control, cache access, to mobility modeling and management, which are critical for the effective operation of PCS networks.Channel holding time is of primary importance in teletraffic analysis for PCS networks. This quantify depends on the user's mobility characterized by cell residence time and the traffic model plotted by call holding time. For convenience and tractability, in the literature channel holding time is commonly investigated by imposing the exponential assumption on the distribution(s) of the cell residence time or (and) the call holding time. Field data and simulation study suggest that the exponential assumption may not be appropriate for PCS networks. In this thesis under a general assumption on the distributions of the call holding time and the cell residence time, the characteristics of new call and handoff call channel holding times are studied. Closed-form analytical formulae for the distributions and expected values of new call and handoff call channel holding times are derived.Location management plays a central role in guaranteeing the effective operation of PCS networks. In current PCS networks, a two-tier system of home location register (HLR) and visitor location register (VLR) databases is commonly used for lo- cation management. To improve the performance of PCS networks, several dynamic location management schemes have been proposed. Among the existing dynamic schemes, the movement-based may be the most practical due to its effectiveness and easy implementation under the framework of current PCS networks. To implement location management in PCS networks, cost analysis is a crucial aspect. However, most of the existing cost analyses for the movement-based scheme are too simple and unavailable for PCS networks with HLR/VLR architecture. One reason is the complexity and difficulty associated with the problem. Li et al and Rodriguez-Dagnino et al challenged this task by considering the HLR/VLR architecture. However, the cost analysis developed by Li et al was carried out under the assumption that both the cell and location area (LA) residence times are exponentially distributed. Rodriguez-Dagnino et al furthered Li et al's work by relaxing the restriction set on the cell residence time. However, both Li et al and Rodriguez-Dagnino et al failed to consider VLR location updates caused by terminal movements between LA's. In this thesis, we desert the restrictions imposed on the distributions of the cell and LA residence times, and propose an analytical model to study the movement-based scheme with HLR/VLR architecture. The issue of VLR location updates caused by movements between LA's is carefully and successfully addressed. Analytical formulae for the costs of HLR and VLR location updates are derived by resorting to complex analysis. Numerical study suggests that the total cost is a convex function of the movement threshold, and is sensitive to the variances of the cell and LA residence times. The result presented in this thesis can serve as a guideline for the system design and implementation of the movement-based scheme for PCS networks.The distribution of the interservice time, which is the time between two consecutive call arrivals served by a portable, is a crucial parameter for modeling and cost analysis of location management in PCS networks. The interservice time differs from the interarrival time due to the busy-line effect. However, in the literature most of the cost analyses for location management were carried out by ignoring the busy-line effect, i.e., by identifying the interservice time with the interarrival time, because so far there lacks analytical results for the interservice time distribution with considering the busy-line effect. In this thesis under a general assumption on the distributions of the call holding time and the interarrival time, analytical models are developed to investigate the busy-line effect upon the interservice time distribution. Closed-form analytical formulae for the interservice time are derived. Numerical study shows that big discrepancies exist between the interarrival time and the interservice time, and that the mean of the interservice time is sensitive to variances of the call holding time and the interarrival time. Based on the analytical results for the interservice time distribution, the impact of busy-line effect on modeling of portable movements in PCS networks is further evaluated. Analytical result for the distribution of the number of cell boundary crossings during the interservice time is obtained. Both analytical results and numerical examples indicate that the busy-line effect has great influence on the distribution of the number of cell boundary crossings during the interservice time. Therefore, any cost analysis for location management without considering the busy-line effect may lead to unreliable results. The results presented in this thesis provide a foundation for modeling location management and portable movements in PCS networks.It is an arduous task to perform call admission control (CAC) for PCS networks supporting multimedia services, which demonstrate inherent heterogeneity in traffic characteristics and quality of service (QoS) requirements. The CAC scheme, thinning scheme, which admits a call with certain probability depending on the current traffic situation in a cell, was first proposed by Fang to handle multiple prioritized services in multimedia PCS networks. In the literature, the blocking probabilities of the thinning scheme were analyzed under the assumptions that call streams with different priorities occupy the same number of channels, and that the average channel holding times of all traffic types are identical. Yet, these oversimplified assumptions may not be appropriate for PCS networks, as mobility patterns and data rates of multimedia services differ from each other, and even new calls and handoff calls belonging to the same traffic type may have different average channel holding times. In this thesis we desert some strict assumptions made in the literature and study the thinning scheme under more general and realistic circumstances. We consider that different service types have arbitrary bandwidth requirements and average channel holding times. Two variations of the thinning scheme are investigated and call blocking performance analysis is carried out using the theory of multi-dimensional Markov birth-death process. Numerical analyses demonstrate that by carefully determining the associated parameter sets, the thinning scheme can assure the QoS demands of multimedia traffic and at the same time optimize system capacity. The thinning scheme can be harnessed to design new CAC strategies for multimedia wireless networks.In wireless data networks such as the WAP system, the cached data may be time-sensitive and thus strong consistency must be maintained, i.e., the data presented to the user at the WAP handset must be the same as that in the origin server. In this thesis, a strongly consistent cached data access algorithm, Probability-Based Callback (pCB in short), is proposed for such networks. In the pCB, upon an update arrival, the action taken by the server is not deterministic; the server can either invalidate the cached data entry in the client or send the updated data entry to the client. The pCB scheme can make good tradeoff between communication cost and access delay, which is extremely difficult for most of the existing cache access schemes. Besides, the pCB scheme possesses excellent universal adaptability and thus can adapt to the inherent heterogeneity of wireless networks and applications. We analytically model the pCB scheme, and derive closed-form analytical formulae for the mean communication cost per data entry access and the mean access delay under a general assumption on distributions of the inter-update and inter-access times. It is demonstrated that the existing Push and Callback schemes are special cases of the pCB scheme.