【作者】 张健；

【导师】 刘磊；

【作者基本信息】 山东大学 ， 软件工程（专业学位）， 2022， 硕士

【摘要】 随着我国经济水平的提高以及互联网科技的发展,各种网络电子设备在人们日常生活中变得随处可见。大量网络敏感应用诸如线上网课、远程会议以及视频点播等也开始逐渐兴起,这与服务质量(Quality of Service,简称QoS)的提出是分不开的。QoS保证了网络流量的高速传输以及各种应用的顺畅运行,按照不同用户的需求合理的调度网络资源。但是这也间接导致了网络规模逐渐增大,众多的通信网络节点相互连通,各种链接更是交错复杂,使得网络拥塞现象频繁发生。网络拥塞具有“传染性”,它会从最初的节点开始,逐渐向外扩散,使其余的相邻节点甚至整个网络产生拥塞。最终将会导致网络的延迟大大增加、丢包频繁甚至瘫痪崩溃。因此,及时识别网络中的拥塞现象,并提供合理的拥塞控制方案是当前亟待解决的问题。本文首先介绍了网络拥塞的成因和危害,详细阐述了国内外目前广泛使用的拥塞识别控制方法以及拥塞控制评价指标。然后概述了 QoS技术领域的基本内容,并对网络流量的自相似等特性进行分析,介绍了每种特性与之关联的流量预测模型。针对网络中节点数量繁多、拥塞识别难度大的问题,提出了一种基于蚁群算法的网络拥塞识别方法。该方法借助改进后的蚁群算法进行寻优,并通过监测拥塞前后最优交叉路径的改变,成功缩小拥塞识别范围,进而确定具体拥塞位置。该方法对解决NP-C问题、快速寻找网络中的拥塞节点,为在根源问题上提升网络通信质量提供了重要帮助。接着本文引入随机早期检测(RED)算法对找到的拥塞节点进行控制,并对RED算法不足之处进行改进,提出了曲线RED即CRED算法,使其更适应网络流量的实际变化情况。最后将CRED算法与蚁群算法相结合,对发现的拥塞节点进行了合理控制,较好地解决了网络拥塞问题。在后续工作中,本文还借助MATLAB和NS2平台,对基于蚁群算法的网络拥塞识别方法进行了仿真,证明了借助蚁群算法判断交叉路径的改变来进行网络拥塞识别不光是简单易行的,而且还避免了传统测量方法带来的一系列弊端。本文还将RED算法和CRED算法进行对比仿真,证明了 CRED算法在拥塞控制方面更加优越。与此同时,还证明了基于CRED算法的拥塞控制方法可以使网络拥塞得到明显的缓解,与蚁群算法相结合后可以快速找到一条满足用户需求的低时延路径。本文最后将基于蚁群算法的网络拥塞识别方法实际应用到城市路网系统中,再次验证了这种拥塞识别方法可以准确找到城市拥挤路口,此外还将基于CRED算法的拥塞控制方法在通信网络系统中模拟,为交警提供了一种缓解交通拥挤的思路。更多还原

【Abstract】 With the improvement of my country’s economic level and the development of Internet technology,various network electronic devices have become ubiquitous in people’s daily life.A large number of network-sensitive applications such as online courses,remote conferencing,and video-on-demand have also begun to emerge gradually,which is inseparable from the proposal of Quality of Service(QoS).QoS ensures the high-speed transmission of network traffic and the smooth operation of various applications,and reasonably schedules network resources according to the needs of different users.However,this also indirectly leads to the gradual increase of the network scale.Numerous communication network nodes are connected to each other,and various links are intertwined and complex,which makes network congestion frequently occur.The network congestion is "contagious",it will start from the initial node and gradually spread outward,causing the remaining adjacent nodes or even the entire network to be congested.Eventually,the network delay will be greatly increased,the packet loss will be frequent,and even the network will collapse.Therefore,it is an urgent problem to identify the congestion phenomenon in the network in time and provide a reasonable congestion control scheme.This thesis first introduces the causes and harms of network congestion,and elaborates the congestion identification and control methods widely used at home and abroad and the evaluation indicators of congestion control.Then it summarizes the basic content of QoS technology field,analyzes the characteristics of network traffic such as self-similarity,and introduces the traffic prediction model associated with each characteristic.Aiming at the problems of the large number of nodes in the network and the difficulty of congestion identification,a network congestion identification method based on ant colony algorithm is proposed.The method uses the improved ant colony algorithm for optimization,and by monitoring the change of the optimal cross path before and after congestion,the congestion identification range is successfully narrowed,and the specific congestion location is determined.This method provides important help for solving the NP-C problem,quickly finding the congested nodes in the network,and improving the quality of network communication on the root problem.Then this thesis introduces the random early detection(RED)algorithm to control the found congested nodes,and improves the shortcomings of the RED algorithm,and proposes the curve RED or CRED algorithm to make it more suitable for the actual changes of network traffic.Finally,the CRED algorithm is combined with the ant colony algorithm to reasonably control the found congested nodes,which can better solve the network congestion problem.In the follow-up work,this thesis also uses MATLAB and NS2 platform to simulate the network congestion identification method based on ant colony algorithm,which proves that using ant colony algorithm to judge the change of cross paths to identify network congestion is not only simple and feasible,and it also avoids a series of drawbacks brought by traditional measurement methods.This thesis also compares and simulates the RED algorithm and the CRED algorithm,which proves that the CRED algorithm is more superior in congestion control.At the same time,it is also proved that the congestion control method based on the CRED algorithm can significantly alleviate the network congestion,and can quickly find a low-latency path that meets the needs of users after being combined with the ant colony algorithm.Finally,this thesis applies the network congestion identification method based on ant colony algorithm to the urban road network system,and once again verifies that this congestion identification method can accurately find the urban congestion intersection.It provides traffic police with a way of alleviating traffic congestion.更多还原