节点文献
移动智能网的过负荷控制机制研究
Study on Mechanism of Overload Control for Mobile Intelligent Network
【作者】 王田;
【导师】 曹长修;
【作者基本信息】 重庆大学 , 控制理论与控制工程, 2002, 博士
【摘要】 智能网以其开放的业务体系结构与移动通信技术紧密结合形成了移动智能网。移动智能网中业务控制的高度集中性带来了业务控制点(SCP)上的高负荷性,特别是在第三代移动通信系统中SCP在整个网络中的地位更加突出,原有移动网的很多功能如切换、位置更新、鉴权等都由SCP负责实现,大量的突发性移动业务极易造成SCP过负荷。对移动智能网的业务特性和过负荷控制机制进行研究具有重要的实际意义,寻求高效、实用的负荷状态检测方法和过负荷控制算法是第三代移动通信系统中必须解决的重要课题之一。本文在吸收国内外有关智能网的业务流量特性和负荷控制研究成果的基础上,针对移动智能网的体系结构特点和业务流量特性,对移动智能网中的负荷状态检测方法和过负荷控制机制及实现技术进行了深入研究,提出了新的负荷状态检测方法和过负荷控制算法。全文的研究结论和主要内容如下:(1)对移动智能网的演进过程进行了研究。随着核心网络的演进和技术体制的变化,智能网和移动网逐步在第三代移动通信系统中完全融为一体。同时,SCP和HLR的功能在同一网络实体中的集成也给节点带来了更高的业务处理负荷。(2)对移动智能网的业务流量特性进行了分析和研究。移动智能网的业务流量具有的自相似性(或称为长时相关性),这种特性对网络性能有显著的影响,在负荷控制中必须考虑业务流的自相似性。业务流的自相似性带来的强突发特性使得表征业务模型的传统描述方法不再适合,应寻求新的建模方法和分析方法,采用分数布朗运动模型可以较好的近似描述这种业务特性,从而为移动智能网的业务流量特性分析奠定了基础。文中对自相似性业务流对缓存容量有限的排队系统的影响进行了分析,给出了表征业务性的Hurst参数的小波估计方法,并对多业务流的合成特性进行了初步研究;文中还提出了业务系统的复杂性分析方法,并提出了描述多业务流合成特性的突发度的定义和分析方法。(3)对移动智能网的负荷检测机制进行了深入研究,包括:负荷源、负荷评价参数、参数与负荷的映射、负荷信息收集策略、参数检测、检测数据处理等一系列相关环节。文中给出了负荷状态与内部资源关系的形式化分析,内部资源消耗率与处理的业务流量有直接关系,可以通过检测系统内部的资源消耗率来进行评估系统负荷状态;在对负荷源进行详细分析的基础上给出了移动智能网各节点的负荷状态评价参数;提出了基于数据融合的负荷状态检测思想,对负荷状态的评价涉及多个参数,根<WP=5>据多个参数进行数据融合后综合评价负荷状态可以降低误报率;同时,对于单节点的负荷状态检测提出了基于冗余信息的负荷检测方法,对于网络负荷状态的检测提出了基于影响权值优化分配的负荷检测方法。(4)对移动智能网的过负荷控制算法、策略和实现技术进行了深入研究。综合预防型控制和反应型控制的优点,提出了基于许可证的两级缓冲Percent和Callgap混合算法,并考虑消息处理的优先级,对缓冲器实现动态门限管理;提出了基于多策略的开放式负荷控制思想,使控制策略从业务处理机制中独立出来,在开放式控制框架下使多种的负荷控制策略融合在一起;针对实际系统的过负荷控制提出了多种实现技术,包括:多处理器间的负荷均衡(提出了具有较低计算复杂度的负荷平衡近似算法,与最优解的差异小于1%)、利用备用机和外围机实现负荷缓冲、I/O数据交互采用多线程处理等。(5)给出了移动智能网在大业务量下的处理能力和性能的测试和实验方案,并对测试数据和实验结果进行了分析。
【Abstract】 Mobile Intelligent Network(MIN) is the close integration of Intelligent Network(IN) with it’s open frame for services and mobile communication technique. In MIN, the concentration of service logic control brings immense service traffic and high load to in Service Control Point(SCP). Especially, the SCP in the third generation mobile communication system is the control core of whole network, must fulfil most control functions ever fulfiled by other entities of mobile network, such as: handover、alter location、authentication etc., which would bring higher load to SCP. The immense burst service traffic usually make SCP into overload state. It is important and realistic to have some thoroughly research on services properties and control mechanism to prevent of overload in MIN. Getting efficient practical load detection algorithms and overload control algorithms are important subjects which must be solved in the third generation mobile communication system.In this dissertation, we give thoroughly study on load detection technique、overload control mechanism and realizing tecnique of MIN with considering architecture feature and service traffic characteristic of mobile intelligent network, and based on some research achievement contributed by international and internal scholars on traffic feature、load control theory and mechanism. Moreover, some new load detection technique and overload control algorithm presented by author are discussed in detail.The results of study and main contents in the paper are given as follows:(1) Giving some analysis and study on mobile intelligent network and its evolutionary trends. With the evolutionary of core network and the development of mobile communication technique, intelligent network would completely integrate with mobile network in the third generation mobile communication system. On the other hand, the function integration of SCP and HLR in one network entity brings greater load to single node.(2) Giving some analysis and study on feature of service traffic in MIN. There exist intrincately self-similar or long-range dependence(LRD,Long Range Dependence) characteristics in the MIN traffic. This character have severe impact on network properties, which must be considered in the load control of MIN. Since the self-similar in traffic would bring immense burst to network, the conventional mathematical definition models are not applicable to network traffic, and some new models and analysis methods should be studied to describe the network traffic. Fractional Brown Motion(FBN) is a good model to approximatively describe that feature, which is basis to analysis traffic property of MIN.<WP=7>The influence of the queues under finite buffers by traffic with self-similar feature is anlayzed, and the method using wavelets to estimate Hurst parameter used to describe the burst of traffic is given reference to some papers, moreover the properties of synthetic multi traffic flow are analyzed simply. In this dissertation, method to analyze complexity of services system and model to analyze traffic in MIN are presented , method to definite and analyze burst of synthetic multi traffic flow is given also. (3) Giving thoroughly reseach on load detection mechanism of MIN, include:load sources、load detection parameters、relationship between parameter and load state、strategy of collecting information、parameter detection、data process method etc. In the dissertation, formal definition and analysis of load and resource is given ; Load state could be estimated by detecting internal resource, since internal resource ocupancy is interrelated to service traffic. A novel idea to detect node or system load state based on fusion of multi detection data is presented; It could get more exactly result by fusioning data from multi parameters, since multi parameters are used to detect load state. And method based on superfluous information to detect load of single node and method based on influence value allocated in optimization way to detect whole network are motioned in detail.(4) Giving
【Key words】 Mobile Intelligent Network; Overload Control; Load detection; Traffic Characteristic; Data Fusion; Load balance;