【作者】 那振宇；

【导师】 郭庆；

【作者基本信息】 哈尔滨工业大学 ， 信息与通信工程， 2010， 博士

【摘要】 骨干拥塞、接入困难和内容分发缓慢是当今地面互联网面临的一系列棘手问题。而基于全球覆盖的、具有远程接入和广播特性的卫星互联网是解决这些难题的有效途径,同时也是未来空天地一体化信息网络的重要组成部分。当前,卫星网络正在向卫星互联网发展,注重高效性,以期提供随时随地的通信与信息服务。但是,设计和实现卫星互联网面临很多技术挑战,首要的就是服务质量(QoS, Quality of Service)问题。一方面,地面互联网的IP QoS问题依然存在于卫星互联网。另一方面,由于卫星通信自身特点又使得卫星互联网的QoS问题需要特殊考虑。因此,分析卫星互联网QoS保障需求,设计高效和易于实现的保障算法是一项重要的研究课题。目前,对卫星互联网服务质量保障的研究主要集中在控制拥塞以及TCP在卫星链路下的适应性问题。然而基于端到端的TCP拥塞控制机制往往无法确知网络内部状态,同时TCP的反馈工作机制又使得各种基于TCP层的改进方法仍不能很好地适应卫星网络环境;此外,切换性和移动性又使得卫星网络中的资源分配变得比地面网络复杂。在分析国内外文献的基础上,本文对卫星互联网服务质量保障方法进行了深入研究,从业务层面入手,着重从网络(IP)层解决拥塞问题,同时改善资源分配方式。这些研究工作完善了卫星互联网QoS保障体系,对未来空天地一体化信息网络QoS保障具有积极意义与参考价值。本文针对卫星互联网QoS保障方法,主要研究了以下几个问题:第一,对卫星网络业务自相似性的研究。考虑到近年来业务自相似性对QoS影响的研究多集中于地面网络,而对卫星网络中自相似性及其影响研究较少这一情况,针对卫星网络特点,研究了自相似业务在卫星网络中的汇聚和传播特性;进而,通过建立地面信关站模型,研究了地面有线网络自相似业务经过信关站进入卫星无线链路的传播特性,并通过仿真进行了验证。此外,仿真并分析了自相似业务对网络节点处队列性能的影响,为后文针对这些影响提出改善QoS的算法奠定基础。第二,对自相似业务量预测的研究。针对卫星互联网中业务普遍存在的自相似性对网络性能产生的不利影响,考虑到自相似业务本身具有可预测性,提出了基于改进卡尔曼滤波的自相似业务量预测算法。针对经典卡尔曼滤波无法求解所提滤波模型这一情况,重新推导了滤波过程。为了提高滤波预测精度,还引入在线噪声估计。所提算法不仅精确度较高,并且与真实流量具有相近的相关结构。第三,对基于业务预测的主动队列管理算法的研究。主动队列管理是IP层的网络拥塞控制方法,而业务量预测对于主动队列管理具有指导意义。鉴于现有主动队列管理算法对估计拥塞有效性不高的情况,提出基于业务量预测的主动队列管理算法。该算法使用改进卡尔曼滤波方法对拥塞进行预测,并结合基于测量的分组丢失率共同做出分组丢弃判决。该算法能动态调节分组丢弃概率,具有队列稳定,链路利用率高的特点。第四,对卫星网络公平服务质量带宽分配算法的研究。针对目前卫星网络的带宽分配算法存在着过度预留和公平性欠佳的问题,提出了卫星网络中公平服务质量的带宽分配算法。该算法使用自适应带宽预留策略,根据效用公平带宽分配准则,保持了用户连接的有效性,保障了用户之间服务的公平性。更多还原

【Abstract】 Currently, Internet is confronted with a series of intractable problems, including backbone congestion, difficult access, slow content delivery, etc. As a significant component of air-space-ground integrated information network in the future, satellite Internet, which has advantages of global coverage, remote access and broadcasting, is an effective solution to above problems. For the purpose of providing ubiquitous communication and information services, satellite network is becoming satellite Internet with high effectiveness. However, there are some technical challenges in designing and realizing satellite Internet. Quality of service (QoS) is the primary problem in these challenges. On one hand, the IP QoS challenges in ground Internet will still exist in satellite Internet. On the other hand, due to inherent characteristics of satellite communication, the QoS issue in satellite Internet has some particularity.Therefore, analyzing QoS requirements and designing efficient and easily implemented QoS guarantee algorithms are crucial points in satellite Internet. At present, most of researches on QoS guarantee in satellite Internet concentrate on controlling congestion and suitability of TCP over satellite links. On one hand, TCP congestion control scheme is based on end-to-end not knowing about internal status in networks. On the other hand, various methods based on TCP layer still cannot completely adapt to satellite links due to the feedback mechanism of TCP. In addition, resource allocation in satellite networks is more complex than ground networks due to handoff and mobility. From the perspective of network traffic, focusing on solving network congestion in network (IP) layer and enhancing resource allocation, this dissertation studies QoS guarantee methods in satellite Internet in-depth by referring to the latest researches. Not only does it perfect the QoS guarantee system of satellite Internet, but also have referential significance for QoS guarantee of the future air-space-ground integrated information network. Aiming at key QoS guarantee algorithms in satellite Internet, the dissertation deals with the following aspects:Firstly, investigation in traffic self-similarity in satellite networks. There are many researches on the effect of self-similarity on QoS of terrestrial networks, but there is little work on traffic self-similarity and its effect of satellite networks. In view of the situation, aggregation and propagation properties of self-similar traffic in satellite networks is analyzed considering characteristics of satellite networks. Then, by building a ground gateway model, the propagation property of self-similar traffic is analyzed and simulated when the traffic from wired ground networks is released to wireless satellite links through the gateway. In addition, the effect of self-similarity on node queue performance is simulated and analyzed for the purpose of presenting algorithms to improve QoS.Secondly, research on self-similar traffic prediction. In the light of the negative effect of ubiquitous self-similarity in satellite Internet traffic on network performances and the predictability in self-similar traffic, an improved Kalman filtering algorithm for predicting self-similar traffic is proposed. As classical Kalman filtering method cannot solve the proposed filtering model, a new filtering process is deduced. In addition, on-line noise estimation is introduced to enhance precision. The algorithm is highly accurate with the correlation structure similar to the real traffic trace.Thirdly, study on active queue management (AQM) based on traffic prediction. AQM is a method of controlling congestion in IP layer, while traffic prediction is significant for AQM. Considering low efficiency of existing AQM algorithms in congestion estimation, an AQM algorithm combined with adaptive traffic prediction is presented. The algorithm estimates congestion using the above improved Kalman filtering. Packet dropping decision is made based on the predicted results and a measurement-based packet loss ratio. It dynamically adjusts packet dropping probability, with advantages of stable queue and high link utilization.Fourthly, work on fair QoS bandwidth allocation algorithm in satellite networks. Excessive reservation and deficient fairness are major drawbacks in existing bandwidth allocation algorithms of satellite networks. According to these flaws, a new bandwidth allocation algorithm based on fair QoS is presented. As the algorithm introduces adaptive bandwidth reservation strategy and fair utility bandwidth allocation rule, it maintains effectiveness of user’s connection, while guarantees fairness of QoS among multiple users.更多还原