Research On Optical Network Resource Allocation For Broadband Streaming Services

IP based data services are explosively growing in our network together with various new types of services. Only optical network technology can provide the infrastructure to meet the bandwidth demand inspired by the current and the future broadband services. Optical network technology has been adopted not only in the backbone networks, but also in the the metro and access networks now. Compared with the effect to the backbone networks, the affect to metro and the access networks made by broadband streaming services are much larger and deeper. Therefore, the design and construction of optical metro and access networks are much more challenging. Various video streaming applications have occupied more and more bandwidth in today's optical metro and access networks. The service model is changed. The QoS (Quality of Service) requirements of streaming applications are more restrict. How to meet the QoS requirements of massive streaming services and to provide high cost efficiency is a new challenge to resource allocation algorithms in optical metro and access networks. This dissertation discusses and investigates the routing and wavelength assignment problem in hierarchical optical metro networks and the wavelength and bandwidth allocation problem in WDM/TDM EPONs (Ethernet Passive Optical Networks) in order to adapt to the requirements of growing streaming traffic.Firstly, we research on the static routing and wavelength assignment problem in metro hierarchical optical networks. We discuss the objective of the static routing and wavelength allocation optimization, analyzing the limitation of current static routing and wavelength assignment algorithms using circle construction method. Then, according to the disadvantages of circle construction method, we propose an offline waveband priority algorithm which targets to maximize the number of waveband paths under the condition of minimum number of wavelengths when all requests are given in advance. The network construction cost can be thus reduced in optical metro networks which has high flux and bandwidth.Secondly, the dynamic routing and wavelength assignment problem in hierarchical optical networks is studied. Based on the analysis of dynamic routing problem, we introduce hierarchical service grooming into integrated routing, and we consider the service grooming at both IP layer and optical layer. We propose a new dynamic integrated routing algorithm in hierarchical optical networks, which can effectively utilize the residual bandwidth of hierarchical OXCs. As an online routing algorithm, our algorithm provides reasonable complexity as O(|W||N|2) and therefore generates quick response to dynamic traffic requests. Simulation results show that our algorithm achieves a comparatively low blocking probability, and when combined with waveband formation process, our algorithm can significantly reduce the node costs of the WDM network.Thirdly, the resource allocation algorithms in WDM/TDM EPONs are investigated. We analyze the advantages and the limitations of several typical wavelength and bandwidth allocation (WBA) algorithms in hybrid WDM/TDM EPONs. Based on these analyses, we present a delta cycle resource allocation algorithm which aims to get adapted to the mass flux and bursty nature of VBR streaming traffic better. Our algorithm includes two parts: the delta cycle algorithm and the enhanced GATE generation mechanism. Simulation results show that our algorithm can significantly decrease the average packet delay and the maximum packet delay of different priority services and improve the network throughput in a WDM/TDM EPON.Fourthly, the traffic prediction is combined with the resource allocation in WDM/TDM EPONs. We use the Quality of Experience (QoE) and QoS requirements of broadband streaming services, and also the industrial standard of IP video streaming delivery evaluation: the media delivery index (MDI) measurements. Under the MPCP (Multi-Point Control Protocol) structure recommended by IEEE 802.3ah Ethernet in the First Mile (EFM) Task Force, the existing resource allocation methods buffer the packets arriving during the waiting period until the next cycle, which introduce extra packet delay. In order to avoid voice and streaming traffic from being deferred to the next cycle, we propose a dynamic wavelength and bandwidth allocation algorithm using adaptive linear prediction. An adaptive linear prediction method is developed to estimate the average arriving rate of VBR traffic in the next waiting period for each ONU, so that the DWBA (dynamic WBA) algorithm can assign wavelength and bandwidth more effectively. Simulation results show that our DWBA method can significantly improve the delay jitter performance according to MDI measurements and can achieve relatively high bandwidth utilization.In the last part of the dissertation, the major contribution and research results are outlined and the future research areas are addressed.