【作者】 黄海；

【导师】 廖建新；

【作者基本信息】 北京邮电大学 ， 计算机应用技术， 2009， 博士

【摘要】 移动网络与互联网世界的结合引起了移动通信领域一场前所未有的业务变革。纷繁多变的业务,不断扩展的价值链以及众多富有竞争力的商业模式都向移动计费系统提出了新的挑战。传统的包月制以及基于时长或流量的粗放式计费已经无法满足移动互联网对数据增值业务的计费需要,人们开始逐渐寻求基于内容价值的精细计费模式,内容计费已经成为3G和下一代网络支撑系统研究中的热点。论文对于内容计费的研究主要集中在内容计费的架构和实时计费关键技术的实现两个部分。对于第一个部分的研究,在已有标准规范的基础之上提出了新的基于策略和计费控制的体系结构；对于第二个部分的研究,分别从基于内容的区分服务、基于内容的负载均衡和多业务共享账户计费三个方面提出了相应的策略和算法,以此解决实时计费系统在内容计费架构下的技术实现问题,并对算法进行了实验验证和性能分析。论文对上述研究进行了详细描述,所取得的主要创新工作归纳如下：(一)针对内容计费的架构研究,在3GPP规范的UMTS(Universal Mobile Telecommunications System)计费架构和IMS(IP Multimedia Subsystem)网络的FBC(Flow Based Charging)架构基础之上,提出了一种增强型基于策略和计费控制(ePCC, enhanced Policy and Charging Control)的内容计费架构。ePCC架构模型拓展了FBC仅支持承载层的策略控制范围,将计费策略控制延伸到了网络的所有层次,并考虑了在线计费和离线计费系统在数据模型上的融合,使得智能数据采集、计费关联、基于价值的计费、融合计费、价值链利益分配等内容计费功能得到了全面的支持。同时,详细描述了新的内容计费架构下实时计费系统的一种设计方式。：二)针对实时内容计费中基于内容的区分服务研究,将各种网元发送给在线计费系统(OCS,Online Charging System)的计费请求,按照不同的请求内容划分成不同类别的队列,并将Internet中的比例时延区分(PDD, Proportional Delay Differentiation)模型应用到了OCS服务器,目标是保证不同类别请求的平均系统响应时延到达预定的比值,从而实现基于内容的相对区分服务。针对PDD模型提出了一种基于系统队列长度和到达率的概率调度(PSQLAR, Probabilistic scheduling based on System Queue Length and Arrival Rate)算法,算法的基本思想是利用系统队列的长度、请求的到达率以及预先分配的时延区分参数作为调度的优先级依据,在调度时采用概率的方式选取需要服务的队列请求。实验结果表明,在考虑服务时延的情况下,PSQLAR性能总是优于一些传统的PDD调度算法,当请求服务时延增加和彼此差别很大时,性能优势相对更大,并且有效地满足了内容计费环境下提出的六点QoS(Quality of Service)要求。(三)针对实时内容计费中基于内容的负载均衡研究,将不同内容的计费请求划分成不同的类别,并将不同类别请求所产生的负载差异性作为负载均衡策略重点考虑的因素,在此基础之上提出了一种最小公倍数能力(LCMC, Least Common Multiple Capacity)负载均衡算法。算法考虑了实时计费集群系统中各个服务器处理能力的差异,各个服务器当前的负载,并特别考虑了不同内容的请求产生的负载差异,采用概率的方式进行请求分发。实验结果表明,在高集群异构性和高请求负载异构性的系统中,LCMC能明显优于一些传统的负载均衡算法,能很好地满足内容计费架构中实时计费集群系统独有的特征。(四)针对实时内容计费中的多业务共享账户计费问题,重点研究同一个账户为同时发生的不同内容的计费请求分配资源的合理性和有效性。提出了一种基于服务质量的资源预留、回收和再分配(3R, resource Reservation, Reclamation, and Redistribution based on QoS)算法,算法在3GPP标准规范的资源预留基础之上,引入了资源回收和再分配流程,并能够根据业务的QoS变化,动态调节这些流程。仿真实验对平均接纳请求数、平均完成请求数以及平均单位请求交互次数三个性能指标进行了检验,其结果表明：在正常QoS水平下,3R算法性能均优于同类型的其它算法；在低QoS水平下,3R算法性能优势更加明显。更多还原

【Abstract】 The combination of the mobile communication networks and the Internet world leads to an unprecedented revolution in mobile services.This is a new challenging task for mobile charging and billing systems with the coming of varied services, expanding value chains, and competitive business models. Traditional extensive methods,such as flat-rate,time-based,and volume-based charging, have been unable to meet new requirements of data value-added services. People begin to seek for intensive charging models based on content and value. Therefore, content-based billing has become an important hotspot in support system researches for 3G and next generation networks.This dissertation mainly focuses on the two parts of the contented-based billing architecture and the key technologies of realtime charging. For the first part, a new policy-based charging billing architecture is proposed on the basis of the standard specifications. For the seconde part, related algorithms are provided to solve the technical problems of realtime charging for the contented-based billing from three aspects, which are content-based differentiated services, content-based load balancing, and account-sharing charging for multi-services. Furthermore, simulation experiments are implemented to verify the correctness and to analyze the performance of these algorithms.The principal contributions of the work presented in this dissertation are as follows:(1) For the research on the content-based billing architecture, an ePCC (enhanced Policy and Charging Control) structre is proposed based on the standard specifications of 3GPP for the UMTS (Universal Mobile Telecommunications System) charging and the FBC (Flow Based Charging) of IMS (IP Multimedia Subsystem).Unlike the FBC which only provides the policy and charging control for bearer level, the ePCC model extends the ability to all levels of the network and considers the data convergence of online charging and offline charging systems, which fully supports the functions of content-based billing, such as intelligent data acquisition, charging correlation, value-based charging, convergent billing, and revenue sharing of value chain. Furthermore, the design of the realtime charging system under the ePCC architecture is elaborated on.(2) For the research on the content-based differentiated services, the charging requests sent to State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications the OCS (Online Charging System) from different network elements are divided into different classes of queues based on the varied contents of requests.The Internet PDD (Proportional Delay Differentiation) model is used in OCS (Online Charging System) servers, which aims to maintain pre-specified mean system response-time delay ratios between different classes of requests, and it can achieve the content-based relative differentiated services. A novel algorithm of PSQLAR (Probabilistic scheduling based on System Queue Length and Arrival Rate) is proposed for the PDD model.The basic idea is that one of the requests from all classes of service queues is chosen with some priority probability which is based on the system queue-length, the request arrival rate, and the pre-assigned delay differentiation parameter. The experimental results show that, in considering request service delays, the PSQLAR algorithm always outperforms some other traditional PDD scheduling ones and the performance advantages are relatively greater when request service delays increase and vary greatly from each other. Furthermore, it can efficiently meet 6 QoS (Quality of Service) requirements proposed in content-based billing.(3) For the research on the content-based load balancing, the charging requests of varied contents are divided different classes and the load differences of each class are taken as the key points of consideration for load balancing strategies. Based on it, a LCMC (Least Common Multiple Capacity) load-balancing algorithm is proposed. This algorithm considers the difference between processing capacity of each server, the run-time load of each server, and especially the difference between various request load, distributing requests by the probability approach. Experimental results show that LCMC significantly outperforms some traditional load-balancing algorithms in the system of high heterogeneous cluster and high heterogeneous request load, and meets the unique characteristics of the OCS cluster in content-based billing.(4) For the research on the account-sharing charging for multi-services, it focuses on the rationality and effectiveness of resource allocation for concurrent charging requests of different contents on the same account. A new algorithm of 3R (resource Reservation, Reclamation, and Redistribution based on QoS) is proposed,which introduces resource reclamation and redistribution processes based on the resource reservation of 3GPP standard specifications and dynamically adjusts them according to the changes of QoS.Three performance indexes are investigated in the simulation experiments, which are average number of accommodated sessions, average number of completed sessions, and average number of iterations per accommodated session. The results show that 3R outperforms the same type of other algorithms in the normal QoS level, arid the performance advantages are much better when the QoS is poor.更多还原