Research On Packet Scheduling In Wireless Networks

As a key scientific research area in 21th Centuary, wireless communication has stepped into the 3rd- generation broadband multimedia networks. The goal of the 3rd- generation wireless communication is that mobile users are able to get video/audio/data service anytime and anywhere. However, with the increasing number of mobile users, the special limit of wireless media will inevitably cause the service bottleneck provided by the networks. Among many elements for quality of service(QoS) provisioning, packet scheduling, whose purpose is to guarantee both fairness and maximal usage of channel, is the most fine-grained element to support QoS provisioning at the base-station for downlink and uplink data flows, therefore it is the kernal component of wireless network system for providing QoS guarantee. At present, the work related to ideal flow scheduling system and packet scheduling algorithm for error-prone wireless channel is a hot topic in the research field of wireless communication.Packet scheduling algorithms originate from the wireline network. Although their application in wireline domain is mature, it cannot be totally suitable for the wireless networks due to the unique characteristics of wireless cellular networks, such as location-dependent and bursty channel errors. Generally only part of data flow can be scheduled along the channel, which is different from the situation in wireline scheduling, therefore the resource reservation and packet scheduling on condition of mobile handoff and channel error in wireless environment are more complicated than in wireline environment.Wireless fair queueing is the basis of wireless packet scheduling algorithm. The basic goal of wireless queuing is to emulate the fluid fair queueing when all flows perceive error-free channels, so it is highly needed that each data flow should be provided bounded delay and seperation. In wireline networks, fair queueing has long been a popular paradigm for providing fairness and bounded delay link access among packet flows over a shared unidirectional link. However, adapting fair queuing to the wireless domain is not a trivial task; the concept of "Fairness" should be redefined due to the unique characteristic of wireless network. In this thesis, the fair service model for wireless networks is built up based on the characterstics of wireless data flow, and the network performance, such as throughput, delay, fairness and compensation is also analysed theoretically. In addition, the key problems in wireless fair queueing are also discussed and a united architecture for wireless fair queueing algorithm is proposed.According to the characterstics of wireless channel and data flow, the mechanism of compensation and degration is further analysed. In terms of the practical channel condition of bursty and location-dependent errors, the data flow are separated and classified into three kinds of states, leading, lagging and in-sync. Essentially, wireless fair queueing swaps the channel allocation between leading flows that lead their ideal error-free services and lagging flows that lag their error-free services over small timescales. Therefore, it seeks to make short error bursts transparent to the user so that long-time throughput guarantees are ensured, at the expense of providing coarser properties for delay, instantaneous fairness and throughput. The scheme is applied in different wireless packet scheduling algorithms, which are analysed qualitatively and quantitiatively in the thesis, based on both theory and experiment.A packet scheduling algorithm, called extended channel state independent wireless fair queueing (CS-WFQ+) is proposed. The idea of improving short-term fairness in WGPS model is fully applied in this algorithm. Each flow is assigned a time-varing fair share of service. While the channel condition becomes less, the shared service is getting more with bounded change. The scheme of start time fair queueing (STFQ) is fully applied in this algorithm in which each flow is assigned a dynamic time-varing shared factor, so...