【作者】 赵峰；

【导师】 刘建明；

【作者基本信息】 中国电力科学研究院 ， 电力系统及其自动化， 2014， 博士

【摘要】 随着公司信息化水平不断提升,一些新的工作模式正在逐渐形成,如视频业务会议、信息配送以及远程教学等,大大提升了工作效率。此外,伴随3G、4G移动通信技术的不断成熟与逐步部署以及音视频编码技术的发展,使很多灵活的业务成为可能,应用诸如移动视频电话、视频音频直播、点播及移动视频监控等。上述业务对海量实时流媒体数据传输产生了广泛的需求。对于大型企业(员工超过万人),其实时流媒体应用不同与传统企业或互联网流媒体应用,具有自己的特点,主要是海量用户的在线接入和实时交互需求同时存在,但在线用户相对稳定,频繁上下线情况较少。另外,通信网络的性能及拓扑的感知性较好,可以更好的支撑流媒体传输。因此,根据大型企业实时流媒体应用的特点,研究实时流媒体传输及调度具有重要意义。国家电网公司拥有员工150万,国家电网公司注册的内网桌面终端60万,外网桌面终端20万,如果同时有超过数万、数十万的员工访问视频服务,面对海量用户的在线接入和实时交互需求,传统的流媒体传输系统面临巨大挑战。海量流媒体数据传输对电力通信网带来巨大冲击,当前的基础设施根本无法满足海量流媒体数据传输的需求,传输的实时性更加难以保证。这使各种实时流媒体传输业务的发展越到瓶颈,在此背景下,以视频会议直播为对象,研究大型企业海量实时流媒体数据的传输调度模式显得异常重要。将传统流媒体传输技术在应用于电力通信网时,存在很多问题：(1)目前存在多种传输技术,它们的特点和资源分配成本大不相同,因此选择何种技术需要深入研究。在应对大规模数据传输时,如果采用C/S模式,服务器承载能力受限,网络带宽存在瓶颈；而CDN技术可以解决带宽问题,但是服务器成本昂贵,维护代价较大；P2P技术虽然可以解决服务器负载问题,但是网络瓶颈没有解决,它会带来巨大的网络传输流量,尤其是带来了大量的跨网域的冗余流量,浪费宝贵的骨干网和网问互联带宽。因此,如何综合利用这些技术,扬长避短,是需要研究的问题。(2)长期以来在传统互联网中视频应用中,视频体验差是一个关键问题,视频启动延迟较大影响着用户的感受,阻碍了流媒体应用的发展。因此,如何降低视频启动延迟,提升系统服务质量(QoS)和用户体验(QoE)已是亟待解决的问题。(3)在电网系统中,终端的网络环境差异很大,有线、无线节点异构节点同时存在,异构节点对视频的数据传播方式带来了挑战,一方面移动终端是否能够满足电力业务需要进行进一步进行实际验证；另外,异构节点对视频质量的要求不一样,带宽性能不一样,视频数据传输调度如何满足不同接入终端的需要,是一个亟待解决的问题。论文基于复杂网络理论、数据分发、视频编解码等术对电力海量数据实时调度及传输进行研究,以电力视频会议直播为突破口,对实时流媒体数据分发的关键性能,包括视频质量、启动及传输延时、跨网流量、负载均衡等进行多目标优化。论文采用两层优化方法,一是对覆盖网拓扑进行优化；二是对数据传输与调度算法进行优化。论文首先对各种数据分发技术进行了全面的分析评价,构建了一种适合电力海量实时流媒体数据分发混合传输架构；对移动终端的接入能力进行验证,论证其应用于电力视频业务的可行性。基于复杂网络理论,搭建了一种基于节点粘度对等网拓扑模型(P2P-NMNV).并针对该模型拓扑度量指标进行了全面分析,结论表明该网络的节点度符合幂率分布,有较短的特征路径和较高的聚集系数,适合实时流媒体数据分发的,有利于降低传输延时,提高用户体验。结合P2P-NMNV模型,在电力系统首次引入P2P-CDN网络模式,并提出了一种基于自治域的混合组网架构HPCS。HPCS综合利用了P2P、CDN和分自治域传输的优点。通过采用分域传输的方式,有效降低骨干网流量,降低传输延时；采用P2P与CDN技术相结合混合组网模式,可以有效提高视频服务质量,提高网络承载能力；采用P2P核心节点与CDN网关互用的模式,大幅的降低设备部署成本。基于熵权分析法,提出了核心节点选举算法,优化了网络传输性能；采用CDN节点单域多核心的模式,核心节点互为备份,提高了网络的鲁棒性,提升了数据传输的效率；为了提高系统容错能力,在核心节点出现故障的情况下,通过采用跨网域连接的方式,提高系统鲁棒性,并得到了跨网流量的最佳配置比例。基于HPCS混合组网架构,构建一种适合于固网和移动混合环境下的视频直播调度模型,并基于熵权层次分析法提出了一种多目标数据调度算法。实验结果表明该算法可以在网络异构的情况下,实现了局部最优的数据调度,在P2P流媒体低延时播放、高视频质量和负载均衡之间达到一个局部最优的配置。综上所述,论文基于复杂网络理论、数据传输技术,在电力海量实时数据传输方面进行了创新,以视频会议直播为突破口,提出了传输模型和数据调度算法,并取得了丰硕的成果,有效解决实时视频数据传输存在的传输瓶颈,并获得了良好的用户体验。本成果适用于各种实时、非实时的海量数据传输应用,在电力系统和互联网领域具有广阔的应用空间。更多还原

【Abstract】 With the development of information technology, the new work mode is gradually formed, such as video meeting, information distribution, and distance education, which greatly enhances the work efficiency. At the same time,with the3G and4G mobile communications technology continues to mature and deployment, as well as the development of audio and video encoding technology, the flexible business applications becomes possible, such as mobile video telephony, video on demand and mobile video surveillance. All of the above needs the transmission of high volume real-time streaming media data.For the large enterprises (more than1million employee), the streaming media application is different with traditional enterprise or internet streaming applications, has its own characteristics, that is massive online access and real-time user interaction needs exist, the online users is stable, the case of online and offline frequently is less. Further, the communication performance and perceptibility of the network is good, can better support streaming media transmission. Considering the characteristics of large enterprise real-time streaming media application, the research of real-time streaming media transmission scheduling is important.The staff of SGCC is more than1500000, the intranet desktop terminal is600000.and the internet desktop terminal is200000. If more than tens of thousands, hundreds of thousands of employees access video service, there will be massive online access needs and real-time interactive demand, the traditional streaming media transmission system faces great challenge. The massive data transmission of streaming media has brought great impact to the electric power communication network.The current infrastructure can’t meet the demand on real-time transmission of high volume streaming data. In SGCC, the live video conferencing system is limited by the capacity of the communication network bandwidth and server. The high volume real-time streaming media transmission becomes bottleneck. So the study of high volume real-time streaming media data transmission is extremely important.There are many problems, when the traditional technology of data transmission used in electric power communication network:(1) There are a variety of transmission technologies, with different characteristics and costs, therefore the choice of distribution technologies should be studied. Deal with large-scale data distribution, using C/S mode, the server load capacity is limited by bottlenecks in network bandwidth; the CDN technology can solve the bandwidth problem, but the server is expensive and larger maintenance cost; P2P technology can solve the server load, but can’t solve the network bottlenecks, it will bring huge network traffic flows, in particular, brought a lot of cross-domain redundant traffic, wasting valuable bandwidth of the backbone network and interconnection network. Therefore, how to use these technologies needs to study.(2) For a long time, in the traditional internet video applications, the poor video experience is a key issue, the long startup delay impact on the user’s feelings, and affect the development of streaming media applications. Therefore, how to reduce startup delays, improve system quality of service (QoS) and quality of experience (QoE) is to be solved.(3) In SGCC, the terminal environment is different, wired and wireless node heterogeneous nodes exist. The heterogeneous node makes challenges on the transmission of the video data. Wheather the mobile terminal is able to meet the requirement of power business or not, needs to be verified. The video quality and the bandwidth requirement of heterogeneous node is different. How the data transmission scheduling meets the needs of different terminal, is an urgent problem to be solved.Based on the theory of complex networks, data transmission and video codec technology, taking power video conference as the breakthrough point, real-time scheduling and transmission of electricity high volume data is studied, aimed to multi-objective optimize the performance of the high volume real time data, including video quality, startup delay, inter-network traffic, load balancing. In this paper, a two-layer optimization method is used, which optimizes the overlay topology firstly and optimize data transmission scheduling algorithm secondly.Firstly, this paper gives a comprehensive analysis and evaluation on a variety of data distribution technology and presents a suitable hybrid transmission architecture for the high volume real-time data distribution. The access capability of the mobile terminal is verified and the feasibility is argued when applied to power video business. Based on the theory of complex networks, this paper presents a peer-to-peer netwok mode based on node viscosity (P2P-NMNV). and analysis the topological metrics for the model, the conclusion show that degree of node in the network is in line with a power-law distribution, short characteristic path and high clustering coefficient, suitable for the power of high volume real-time data distribution, reducing transmission delay and improve user experience.Based on the P2P-NMNV model, a hybrid network architecture based on autonomous domain HPCS (Hybrid P2P and CDN Structure) is present, with comprehensive utilization of the advantages of P2P, CDN and sub-autonomous domain transfer. This is the first time for using P2P-CDN network model in the power system. For Reducing backbone traffic and delay by transmission in different area effectively; the hybrid network mode, combined P2P and CDN technology, can effectively improve the video quality of service, and improve network carrying capacity; with the P2P core nodes CDN network exchangeable mode significantly reduce the equipment deployment costs. Using the analysis method based on entropy weight, present core node election algorithm to optimize the network transmission performance; With single domain multi-core CDN node, the core nodes connected with each other, it can improve the robustness of the network and enhance the efficiency of data transmission; To improve the system fault tolerance in case of failure of the core node, the cross-domain connections can be used, which can improve the system robustness, and get the inter-network traffic allocation ratio. Depend on the live streaming data scheduling model based the HPCS framework, a scheduling model is present, which is suitable for a mixed broadcast environment of fixed and mobile video. A multi-target data scheduling algorithm based on entropy weight analysis method is proposed. The simulation results show that the algorithm can achieve the local optimal data scheduling algorithm, which can get a local optimum configuration between low-latency playback, high video quality and load balancing.In summary, based on the complex network theory, Data transmission technology, taking the video conference as the breakthrough point, an innovation is proposed in the power high volume real-time data transmission model and data scheduling algorithms, which achieves fruitful results, effective solution to a real-time video data transmission, and a good user experience. The achievement is suitable to the real-time, non-real-time mass data transmission applications in power systems and the internet.更多还原