Research On Dynamic Scheduling And Cross-layer Optimization On Broadband Wireless Networks

In recent years, the rapid development of wireless network leads to the integration of multiple services (i.e. data, voice, message transmission and video on demand), and the next-generation network is expected to provide different QoS (Quality of Service) for such integrated services. Scheduling algorithm is the key part for QoS ensurance in wireless broadband network, especially for the real time services. However, due to the long time research history and the abundant results, the capability of individually relying on the network to improve the performance of scheduling algorithm for further efficiency improvement is limited. As the key technology for the future communications, cross-layer optimization has broken the strict boundaries between the layers in the traditional open systems interconnection (OSI) model, so that the relevant information of different layers can be shared with each other. Cross-layer optimization can coordinate and integrate the characteristic parameters which are scattered in different layers and make the stacks be able to adapt the changes of QoS demands and network status. Therefore jointly cross-layer optimization of the scheduling algorithm in wireless broadband network for the effective resources management can provide a broader space for the wireless broadband network. It is important to effectively improve the QoS performance.The theoretical analysis on scheduling algorithm is always complicated, even the simplest algorithm. But the precise theoretical analysis is an effecitive way to analyze the performance of scheduling. Theoretical analysis is also the basis for the performance improvement and the algorithm implementation. Actually, in TDMA system (Time Divided Multiple Access) system, a problem is liable to be ignored: the delay is calculated and expressed in the unit of slot. When the packets arrive at the stations in the same slots, although their access moment is different, the access slot ordinal is the same. Hence, it is not proper to calculate delay in the unit of time (seconds, milliseconds, etc.) directly.This dissertation investigates in-depth on the scheduling algorithm and its cross-layer optimization in terms of theoretical analysis and practical application. The research includes the dynimc scheduling algorithm and its cross-layer optimization under symmetrical network environment (the network parameters, such traffic arrival process, serve rate, etc., are the same for different stations), the dynamic scheduling algorithm and its cross-layer optimization under asymmetrical network (the network parameters, such traffic arrival process, serve rate etc. are not the same for different stations). Based on the analysis of polling-based scheduling algorithm, the first part proposes a novel dynamic scheduling algorithm, M-gated scheduling. By the method of Probability Generation Function (PGF), the mathematical expressions of queue length in the buffer, packets delay and polling cycle period are obtained, which are validated by the simulation results. Compared with other similar algorithms under the same simulation environments, the performance of the proposed algorithm is close to that of the optimal scheduling algorim. Moreover, the unfairness can be avoided. In the meanwhile, M-gated also performs better in terms of system stability, flexibility and robustness.Then a new cross-layer designing combined with M-gated scheduling and admission control is proposed and used in IEEE 802.11 wireless local access network. Compared with other designing, the network is more stable, the parameter overheads are less and the delay QoS can be maintained in an acceptable level. Moreover, this designing improves the performance of protocol while the standard structure is not changed much. Therefore the proposed designing is especially suitable for the heavy loads and the burst traffics.However, in the real network environment, the varieties of traffics bring about the differentiation in delay, error bit and transmission rate, which should not be overlooked. Consequently, the dissertation extendes the results obtained under the symmetric networks and investigates the asymmetric M-gated scheduling and its performance.The first difficulty is that under the asymmetric network environment, the number of variables which denote the arrival rate, serve rate and switch-over period increases N-fold (N stand for the number of users), which makes the theoretical analysis for M-gated very complex. Among these difficulties, the analysis for the second moment of the queue length is the hardest task. So the theoretical calculations for asymmetric M-gated scheduling take two steps: First, we focuse on asymmetric 1-gated and asymmetric 2-gated to acquire the intermediate variables which are helpful for the theoretical analysis for the asymmetric M-gated scheduling, especially for the second moment of queue length; second, we bring into these variables into the theoretical analysis for the asymmetric M-gated scheduling to acquire the complete expressions about the multi-PGF, average queue length, average delay and average cycle time.Extensive simulations validate the mathematical analysis of asymmetric M-gated scheduling and many special features which are caused by asymmetric network environment are also observed.At last, based on those special features, a cross-layer design is proposed in IEEE 802.16 networks after fully considered the virtue of wireless channel and the link adaptive technology. The designing adds three modules, i.e. admission control module, scheduling module, buffer module and central parameters controlling module, in IEEE 802.16 networks to fully use the network resources and ensure the QoS demands of rtPS, nrtPS and BE traffics. Simulations show the cross-layer designing can significantly improve the QoS of various service types and do a better compromise between bandwidth usage, parameters transmission overheads and surplus bandwidth.In brief, the thesis focuses on the dynamic scheduling algorithm and its cross-layer design. After fully considering the feature of TDMA system, the mathematical models are built. By extensive theory analysis and simulations, the performances of dynamic scheduling under symmetric and asymmetric network environment are investigated. Furthermore, based on these results, the cross-layer designs combined with network layer and physical layer are proposed and used in the practical wireless network. Extensive simulation results show the effectiveness on improving the network performances.