认知网络中基于流量预测的负载均衡技术研究和实现

发布时间：2018-02-09 23:08

本文关键词： 认知网络 QoS 流量预测负载均衡　出处：《北京邮电大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着未来互联网的架构及使用行为模式所发生的翻天覆地的变化,网络中的数据流量也将随用户数量、服务应用类型的增多而快速增长。对于日趋复杂的网络接入环境,如何根据网络环境状态进行有效地网络流量预测与负载均衡决策,更好地协调各种网络资源,解决网络拥塞和服务质量下降等一系列问题,对于未来异构网络流量控制与性能测量具有重要的意义。认知网络不仅具有频谱动态感知切换功能,还具有对节点地理信息、链路质量、网络流量、用户偏好、业务类型以及服务质量等方面的认知学习决策能力,因此可以更好的进行网络流量预测,进一步改善网间互通,优化网络流量,实现负载均衡,提升网络性能。本文通过对传统互联网流量预测技术的研究和认知网络智能感知特性的分析,并重点结合认知网络中的流量预测与负载均衡技术,开展本课题工作,主要工作如下：首先研究了当前网络中的一些流量预测方法,对其中的ARIMA模型及马尔科夫链模型原理进行了详细的分析,并利用仿真实验对这些预测模型进行了实验。本文在此基础上提出了一种基于部分可观马尔科夫过程的认知网络流量预测模型。该方法充分考虑到了认知网络中多种参数的不可测性,不完整性,可以根据认知网络中前一时刻的状态,采用部分可观测的参数进行流量预测,解决流量预测中的不可观参数选择问题,提高流量预测精度。然后本文根据上述的流量预测模型提出了一种认知网络负载均衡算法,可根据所设计的分类方法及预测算法对传输业务分类和预测结果对网络流量进行合理调度和均衡,提高网络吞吐量,改进网络资源利用率与数据传输性能。最后,本文在认知网络系统中进行该课题的实现,对传统模式下的网络状态与采用本方案下的网络状态进行比较,选取网络吞吐量,丢包率,时延,抖动等参数进行了具体分析,体现出了本方案的优势。
[Abstract]:With the great changes in the architecture of the Internet and the mode of using the Internet in the future, the data flow in the network will also grow rapidly with the increase of the number of users and the types of service applications. How to make effective network traffic prediction and load balancing decision according to the network environment condition, better coordinate all kinds of network resources, solve a series of problems such as network congestion and quality of service decline, etc. Cognitive network not only has the function of spectrum dynamic sensing switching, but also has the function of node geographic information, link quality, network traffic and user preference. Because of the cognitive learning decision ability of service type and quality of service, it can better predict network traffic, further improve interworking, optimize network traffic, realize load balance, and improve network performance. Through the research of traditional Internet traffic prediction technology and the analysis of cognitive network intelligent perception characteristic, and combining the traffic prediction and load balancing technology in cognitive network, this paper carries out the work of this subject, the main work is as follows:. Firstly, some current network traffic forecasting methods are studied, and the principle of ARIMA model and Markov chain model are analyzed in detail. On the basis of these experiments, a traffic prediction model of cognitive network based on partially observable Markov process is proposed. This method takes full account of the cognitive network. The untestability of multiple parameters, According to the state of the previous moment in the cognitive network, the incomplete state can be used to predict the flow with some observable parameters, which can solve the problem of the unobservable parameter selection in the traffic prediction and improve the accuracy of the traffic prediction. Then, according to the above traffic prediction model, this paper proposes a cognitive network load balancing algorithm, which can reasonably schedule and balance the network traffic according to the designed classification method and prediction algorithm. Improve network throughput, network resource utilization and data transmission performance. Finally, this paper carries on the realization in the cognitive network system, compares the network state under the traditional mode with the network state under this scheme, selects the network throughput, the packet loss rate, the delay, The jitter and other parameters are analyzed in detail, which shows the advantages of this scheme.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.06

【参考文献】