卫星互联网自相似业务预测方法的研究

发布时间：2018-11-20 13:44

【摘要】：地面因特网规模的空前繁荣推动了下一代网络(Next Generation Network)的迅猛发展，形成了集合陆、海、空、天的空天地一体化信息网络。作为地面网络的旁路网络，宽带卫星互联网缓解地面互联网的压力的同时，又面临着一系列亟待解决的困难。如宽带卫星互联网以TCP/IP为承载方式和信息传输平台，使得对其原本适用于地面的网络协议的实用性有了新的要求；而地面-卫星的混合网络本身的复杂性以及对互融网络资源的分配也有较大的难度等问题使得对卫星互联网服务质量(QoS，Quality of Service)保障的讨论成为一项必要且重要的研究课题。着眼于宽带卫星互联网QoS的要求，本论文从卫星节点接收和转发分组数据的角度，分别论证星间骨干网中自相似业务量通过一次跳变在目的卫星节点处汇聚的流量的自相似特性，以及卫星作为传输数据的中间节点转发数据的自相似特性。另外，还探讨了地面-卫星混合网络中从地面信关站转发到卫星链路上的流量数据的自相似性；并通过建立ON/OFF源的自相似业务流量模型以及Pareto/M/1/C的排队模型来实验证明卫星通信系统中传输的数据具有自相似性。最后，论证了卫星网络中业务的自相似程度的增加能导致更高的系统丢包率和系统所需带宽，从而使通信系统的性能恶化，降低QoS。由于卫星通信系统上的流量不可获取，因此通过对卫星通信系统中数据传输模型的分析，采用适当的地面网络自相似流量作为参考来建立基于小波的ARIMA预测模型。在理论指导下通过Matlab仿真实验，选取基础小波sym小波对非平稳、周期变换且具有长相关性的网络流进行三层小波分解，从而使自相似网络流量平稳化且使其自相似程度得以降低。针对本论文提出的预测模型，选取预测样本数目为100，预测步长为5，预测样本数与预测数的比为20:1的预测模型对小波分解后的数据进行预测能够达到最好的预测效果，其预测均方根误差为0.1773。本系统的预测结果用于提供卫星星上调度的队列信息、为卫星互联网的跨层调度以及带宽分配等提供可靠的参照数据，从而为实现网络拥塞的缓解、优化资源分配奠定了基础，进而改善卫星网络性能，，提高卫星网络传输的服务质量。
[Abstract]:The unprecedented prosperity of the terrestrial Internet promotes the rapid development of the next generation network (Next Generation Network), and forms an integrated information network of land, sea, air and sky. As a bypass network of terrestrial network, broadband satellite Internet is facing a series of difficulties to be solved while relieving the pressure of terrestrial Internet. For example, the broadband satellite Internet takes TCP/IP as the bearing mode and the information transmission platform, which makes the practicability of the network protocol which originally applied to the ground has new requirements. However, the complexity of the ground-satellite hybrid network and the difficulty of resource allocation in the intermingled network make it possible to improve the quality of service (QoS,) of the satellite Internet. The discussion of Quality of Service) guarantee has become a necessary and important research topic. Focusing on the requirements of broadband satellite Internet QoS, this paper from the point of view of receiving and forwarding packet data from satellite nodes, The self-similar characteristics of self-similar traffic converging at the destination satellite node by a single jump and the self-similar characteristics of the satellite transmitting data as the intermediate node of the transmission data are discussed respectively in the inter-satellite backbone network. In addition, the self-similarity of traffic data transmitted from ground to satellite link in ground-satellite hybrid network is discussed. The self-similar traffic model of ON/OFF source and the queuing model of Pareto/M/1/C are established to prove that the data transmitted in satellite communication system has self-similarity. Finally, it is demonstrated that the increase of service self-similarity in satellite network can lead to higher packet loss rate and bandwidth required by the system, thus worsening the performance of the communication system and reducing the QoS.. Because the traffic on the satellite communication system is not available, the ARIMA prediction model based on wavelet is established by analyzing the data transmission model in the satellite communication system and using the proper self-similar traffic of the ground network as the reference. Under the guidance of theory, through Matlab simulation experiment, the basic wavelet sym wavelet is selected to decompose the non-stationary, periodic and long-correlation network flow with three-layer wavelet decomposition. Thus the self-similar network traffic is stabilized and its self-similarity degree is reduced. According to the prediction model proposed in this paper, the prediction model with the number of prediction samples is 100, the prediction step is 5, and the ratio of prediction sample number to prediction number is 20:1, the best prediction effect can be achieved by the prediction model after wavelet decomposition. The root mean square error of prediction is 0.1773. The prediction results of this system are used to provide the queue information of satellite on-board scheduling, and to provide reliable reference data for the cross-layer scheduling and bandwidth allocation of satellite Internet, thus laying a foundation for the alleviation of network congestion and the optimization of resource allocation. Furthermore, the performance of satellite network is improved and the service quality of satellite network transmission is improved.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN927.2

【相似文献】