Q O S OPTIMIZATION THROUGH CAPACITY AGGREGATION OF MULTIPLE LINKS IN HETEROGENEOUS WIRELESS NETWORKS

Abstract

Wireless and mobile networking is considered as the most appreciated technologi- cal innovation that has stormed into the life-styles of the people and has found applica- tions in business, education, health, social networking and all other major areas of day-to- day work. Inspired by the enormous potential of mobile computing, wireless communica- tion and mobile networking technologies have emerged as major research disciplines within the domain of communication systems. Most of the communication based applica- tions have been developed and executed on robust wired communication networks and when they are ported to mobile devices, face serious challenges in meeting quality of ser- vice (QoS) requirements over the mutable wireless media. The QoS enabling for mobile networks has, therefore remained a major research area in this discipline. The presence of multiple communication interfaces in modern multi-mode mobile devices has enabled a new dimension for improving QoS solution during mobility but the heterogeneity of these networks pose serious challenges for delay bounded QoS aware applications. In this the- sis, capacity aggregation (CAG) of multiple available wireless links has been investigated to quantify its suitability in providing QoS during mobility. There are numerous challenges faced by the designers and developers in synchro- nizing flows, transported over multi-path. One such issue is the out-of-sequence (OOS) reception of packet at the receiver due to transport of these packets on multiple heteroge- neous paths. The larger extent of OOS receptions; generally cause serious performance degradation in delay bounded real-time applications, and is also a source of expensive buffer management for in-order delivery of packets to their respective destinations. The problem gets adverse during mobility, as the end-to-end path repeatedly changes during mobility and accompanies with changes in characteristics of E2E path that enhances probability of OOS reception. Therefore, maximization of QoS through minimization of OOS reception in a CAG environment is the main problem investigated in this thesis. The problem has been persuaded with novel multi-server scheduling schemes that minimize the OOS reception at the receiver. The proposed multi-server scheduling schemes provide a general service model that accommodates multiple types of traffic flows belonging to different classes-of-service. This approach is in contrast to the QoS maximization of a single flow at upper layers of TCP/IP protocol stack. The approach of multi-server scheduling to minimize OOS reception has been in- vestigated with the help of novel deterministic and stochastic analytical models to provide quantified solution of the above given problem. The fundamental mathematical structure of novel deterministic and stochastic models has been elaborated to enumerate multi-path transport dynamics of a mobile flow. These models provide appropriate formulae for the dimensioning questions of the QoS during mobility. Hence, the analysis of the multi-path transport of mobile flows remains as the main focus of this research work and constitutes core of this dissertation. Since the deterministic and stochastic framework for the analysis of E2E multi-path dynamic are not fully developed, my research focused on efficient es- timations of E2E delay variations and its impact on OOS reception of packets. The tight deterministic and stochastic performance bounds have been derived through proposed models. The main application of these models has been exercised in development of suit- able scheduling strategies that minimize E2E delay variation and OOS reception in CAG during mobility. The results of proposed models have been validated through simulations ‐ v ‐ as well. The results have shown robust performance of proposed scheduling schemes in achieving acceptable QoS levels for real-time flows during mobility, with minimized OOS reception and reduced buffer occupancy.