【作者】 尤天舒；

【导师】 李文辉；

【作者基本信息】 吉林大学 ， 计算机应用技术， 2014， 博士

【摘要】 随着网络技术的不断发展，越来越多的人通过网络来了解信息，由于用户群体的不断增加及网络应用的丰富，使得服务器接收到的请求数越来越多，有关于网络负载的问题越发突出。集群技术的出现很好地解决了这一问题。本文主要贡献及创新点包括：1、提出了一种基于Agent的动态自适应集群负载均衡模型。通过把Agent技术引入集群负载均衡系统中，创新性的构建了一个基于Agent的动态自适应集群负载均衡的普适模型。将Agent技术引入到传统集群负载均衡模型中，增加了集群负载系统的智能性和自适应能力；2、创新性的把请求任务识别与分类思想融合到了基于Agent的集群动态负载均衡模型中。通过引入请求任务识别与分类思想，结合Agent技术，实现了集群任务的自动识别及分类处理，提高了集群系统的运行效率；3、针对本文提出的基于Agent的集群动态负载均衡模型，对其具体功能设计了详细的算法，并针对模型多个指标做了性能分析。4、提出并实现了一种适用于流媒体业务的基于Agent的动态自适应负载均衡集群。在详细分析了流媒体业务存在的问题及特点的基础上，根据这些问题及特点做了针对性的优化，设计了针对于流媒体业务的集群模型并给出了具体实现。实验结果验证了本文模型的普适性和有效性。5、提出并实现了一种针对移动互联网业务的基于Agent的动态自适应负载均衡集群。针对移动互联网特点及新的挑战，设计了一种针对移动互联网的基于Agent的负载均衡集群，此集群结合本文提出的模型进行设计，同时又兼顾了移动互联网的特点。进一步验证了本文模型的普适性和有效性的同时。更多还原

【Abstract】 With the rapid development of Internet technology, more and more people begin toretrieve news information and do shopping through the Internet. Many people would like tohave an e-account, even many enterprise make transactions through the Internet in order togain more ratification. But troubles are coming. With the increasing scale of Internet users,the load of the Internet is becoming heavier. Although enhancing the performance of theInternet hardware could somewhat relieve this problem，but the continuously increasingaccess request brings even more load to the servers. So, the development of hardware isimportant indeed, but there is also a need of corresponding load balancing strategy.Fortunately, the Computer Cluster technology preferably solved this problem.Cluster (also called Computer Cluster) accomplish computational task by a group ofloose combined software and hardware who cooperate highly tightly. One of the advantagesof cluster technology is that it could achieve high performance, high fault tolerance and highflexibility with a very low cost. Actually, a large scale cluster could be seen as a whole. Eachserver corresponds to a node, and server could connect to each other in number of ways.Reasonable scheduling among tasks is the core of the cluster technology.Nowadays dynamic balancing load technology is widely used in Internet clustertechnology. The load balancing techniques are often divided into two categories: Dynamicload balance and Static load balance. In order to achieve improved performance of the entirecluster system to enhance the overall network load capacity, improve network response timegoal, this paper, a more flexible intelligent Agent technology, this paper proposes a dynamicadaptive load balancing model based on Agent. And through the different request packetanalysis tasks, divided request task into N categories, each category corresponding to arelatively efficient load balancing schedule algorithm, which have enhanced the performanceof the cluster significantly.This paper mainly accomplished the following works:1.Built an Agent based dynamic adaptive cluster load balancing generalized model byintroducing the Agent technology, which could not only improve the intellectuality of the load balancing system, but also enhance the adaptive capability;2.By introducing the idea of recognition and classification of the request task andcombining the Agent technology, achieved the automatic recognition and classification of thecluster task, which further enhance the operational efficiency of the system;3. Based on the aforementioned model, designed a corresponding algorithm andimplement it. Our model is proven to be effective through the comparison with the originalLVS cluster on several performance parameters;4. Analyzed the special characteristics of the streaming media service and its demands onthe streaming media. On this basis, applied the Agent based dynamic adaptive load balancingmodel to the cluster environment of streaming media servers, and do specific optimizationsaiming at the special characteristics of the streaming media;5. Applied the Agent based dynamic adaptive load balancing model to the clusterenvironment of streaming media servers, and the model is proven to be effective in astreaming media server cluster through the comparison with the original cluster;6. Analyzed in depth the characteristics of the mobile Internet itself, including thedifferences of the user’s behavior, different network throughput, service scale, and servicerequirements. And new challenges brought by these were also analyzed;7. Aiming at the characteristics and new challenges, designed an Agent based dynamicadaptive load balancing cluster for the mobile Internet. The cluster is designed according tothe model we proposed in this paper, at the same time give considerations to thecharacteristics of mobile Internet;8. Constructed the model mentioned above in commercial environment, designed andimplemented4service types to access the cluster. The network throughput, service load andservice demands of these service types are different, which almost cover the common servicetypes of mobile Internet, simple and generalized, could validate our work convectively.In summary, this paper introduce the technology of Agent and the idea of recognitionand classification of the request task,presents an Agent based dynamic adaptive loadbalancing model and analyze the performance of this model; innovation used this model instreaming media, do specific optimizations aiming at the special characteristics and demandsof the streaming media; systematically analyses the characteristics and demands of mobileInternet and do specific optimizations on our model according to these characteristics,implement and validate it in commercial environment. The results showed that the theoreticalsignificance and application value of our Agent based dynamic adaptive load balancing modelcan provide a reference to the same kind of research and application.更多还原