基于业务分类的服务器选择算法的研究

发布时间：2018-03-30 21:15

本文选题：内容分发网络　切入点：内容路由　出处：《北京邮电大学》2014年硕士论文

【摘要】：互联网的飞速发展使人们的工作和生活变得越来越便利。然而不断丰富的应用使得网络中的流量呈现爆炸性增长的态势,网络尤其是骨干网的拥堵已经变成了一个必须面对的重要问题。为了解决这个问题,互联网创始人之一Tim berners-Lee于1995年发起了一项解决网络拥塞的技术挑战,这项技术挑战最终促使内容分发网络(Content Delivery Network, CDN)的诞生。CDN从根本上改变了互联网的传输模式,通过在靠近用户的边缘服务域布置大量的服务节点,使得用户能够从距离较近的边缘服务节点访问所需的资源,从而有效缓解了骨干网络的拥塞,同时也改善了用户体验质量(Quality Of Experience,QoE)。服务器选择是CDN中内容路由关键技术需要考虑的重要问题之一,其算法将决定内容路由的优劣,进而影响用户的QoE。传统的服务器选择算法采用轮询、随机、或是最小负载等算法。然而这些算法当突发热点事件时,很容易导致单个服务器负载过重,而且无法适应多业务的应用场景。这些存在的问题将会极大程度的限制CDN的进一步发展与应用。为了解决这一问题,本文在混合式CDN架构基础之上,提出了一种基于虚拟负载的概率重定向算法。在服务域中,每个CDN节点有一个连接队列,能够将到来的连接请求存储在队列中。负载均衡器采用轮询的方式将请求分发到各个CDN节点上,当队列的连接负荷达到规定阈值时,启用本文提出的算法,通过一种概率的方式将本身过载的请求重定向到负载较轻的邻居节点上。该算法能够避免单个节点负载过重的状况发生,充分利用了各个节点的硬件资源来为用户提供更好的QoE。最后搭建了具体的实验平台,并通过Apache Jmeter来模拟多用户并发的场景,分别从用户平均响应时间及节点虚拟负载两方面对本文提出的算法和传统算法进行了比较,实验结果表明了本文提出的算法在平均响应时间上要比经典的轮询和最小负载算法提高至少10%,而且各个节点的虚拟负载相对更加均衡。
[Abstract]:The rapid development of the Internet makes people's work and life more and more convenient. Network congestion, especially in backbone networks, has become an important problem that must be faced. In order to solve this problem, Tim berners-Lee, one of the founders of the Internet, launched a technical challenge to solve network congestion in 1995. This technical challenge ultimately led to the birth of the content Delivery Network (CDNs), which fundamentally changed the mode of transmission of the Internet by placing a large number of service nodes in the edge service domain near the user. It enables users to access the required resources from close edge service nodes, thus effectively reducing the congestion of backbone networks and improving the quality of user experience quality of experience QoS. Server selection is one of the most important issues in content routing in CDN. Its algorithm will determine the quality of content routing, and then affect the QoS of users. Traditional server selection algorithm adopts polling and randomness. Or algorithms such as minimum load. However, when a hot event occurs, these algorithms can easily cause a single server to be overloaded. These problems will greatly limit the further development and application of CDN. In order to solve this problem, this paper proposes a probabilistic redirection algorithm based on virtual load based on hybrid CDN architecture. In the service domain, each CDN node has a connection queue. The load balancer distributes the request to each CDN node by polling. When the connection load of the queue reaches the specified threshold, the algorithm proposed in this paper is enabled. The overloaded request is redirected to the lighter neighbor node in a probabilistic way. The algorithm can avoid the overload of a single node. The hardware resources of each node are fully utilized to provide users with better QoS. Finally, a concrete experimental platform is built, and the multi-user concurrent scenario is simulated by Apache Jmeter. This paper compares the proposed algorithm with the traditional algorithm in terms of the average response time of the user and the virtual load of the node. The experimental results show that the average response time of the proposed algorithm is at least 10 higher than that of the classical polling and minimum load algorithms, and the virtual load of each node is relatively more balanced.
【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

【共引文献】