P2P视频点播拥塞控制及节点选择优化策略研究

发布时间：2018-02-12 07:36

本文关键词： 拥塞控制负载均衡节点选择 Q-学习非结构化P2P网络 VOD　出处：《江苏大学》2017年硕士论文　论文类型：学位论文

【摘要】：视频点播(Video On Demand)是一种交互式视频播放与传输系统,可以满足用户根据自身需求任意观看视频节目。将P2P技术与视频点播技术相融合,可以充分利用节点的计算和存储能力来减轻服务器的压力。但也存在网络负载失衡以及节点选择失效等问题导致视频缓冲等待时间过长,从而引发的用户体验不佳。因此研究网络负载均衡策略以及节点选择方法可以有效地减少用户视频点播中的缓冲等待时间,保证了视频播放的流畅性。为此本文的主要工作如下:1.介绍了相关理论及技术。首先介绍了P2P技术的概念与分类,并将P2P模式与C/S模式作比较,讲解了P2P网络模型分类以及各类型的特点。其次介绍了负载均衡技术的概念及分类,同时讲解了常用的负载均衡方法。2.提出了基于拥塞控制的非结构化P2P网络的路由负载均衡策略。为解决P2P网络节点拥塞导致网络负载失衡的问题,本文提出了一种新的基于拥塞控制的路由查询方法来实现动态网络下的资源查询。该策略分为两部分实现:一是采用资源分组和节点重连策略,将拥有相同资源的节点分到同一资源组内,使网络中组内节点资源定位更迅速且不同资源组间的查询得到了均衡;二是在路由负载均衡中采用一般运用于状态学习的协同Q-学习方法,使得同一组节点上的资源查询被有目的地引导以避开那些组内拥塞的节点,从而最终实现资源组内节点之间的查询均衡。实验结果表明本文所提策略使得资源组间与组内的查询转发有效地避开拥塞节点从而使得网络负载得到了均衡。3.提出了基于线性规划和强化学习的P2P视频点播节点最优选择策略。针对P2P视频点播系统中如何选择节点以充分利用P2P网络资源、减少服务器带宽消耗,本文提出了基于线性规划和强化学习的P2P视频点播节点最优选择策略。该策略分为两层来实现:第一层节点选择策略中服务器利用定期从网络节点采集的全局信息进行线性规划,从而引导节点选择过程;第二层节点选择策略采用协同Q-学习方法,考虑邻居节点的本地信息如剩余上行带宽、资源的个数以及拥塞状态等,在完全分布式环境下寻找能拥有资源且能提供剩余带宽的节点。实验结果表明本文所提策略能够有效地选择合适资源节点集提供资源,同时有效地降低服务器带宽消耗、减少服务器计算量。
[Abstract]:VOD Video on demand (VOD) is an interactive video playback and transmission system, which can meet the needs of users to watch video programs. It can make full use of the computing and storage capacity of the node to reduce the pressure of the server, but there are also some problems such as network load imbalance and node selection failure, which lead to the long waiting time of video buffering. Therefore, the research of network load balancing strategy and node selection method can effectively reduce the buffer waiting time in VOD. The main work of this paper is as follows: 1. The related theories and technologies are introduced. Firstly, the concept and classification of P2P technology are introduced, and the comparison between P2P mode and C / S mode is made. The classification of P2P network model and the characteristics of each type are explained. Secondly, the concept and classification of load balancing technology are introduced. In order to solve the problem of network load imbalance caused by congestion in P2P network, the routing load balancing strategy of unstructured P2P network based on congestion control is proposed. In this paper, a new routing query method based on congestion control is proposed to realize resource query in dynamic network. The strategy is divided into two parts: one is resource grouping and node reconnection strategy. The nodes with the same resource are divided into the same resource group, so that the node resource location in the network is faster and the query among different resource groups is balanced. Secondly, in routing load balancing, cooperative Q- learning method, which is generally applied to state learning, is used to guide resource queries on the same group of nodes purposefully to avoid the congested nodes in the group. The experimental results show that the strategy proposed in this paper can effectively avoid the congestion nodes and make the network load balanced. 3. Based on linear programming and reinforcement learning, the optimal selection strategy of P2P VOD nodes is proposed, aiming at how to select nodes in P2P VOD systems to make full use of P2P network resources. Reduce server bandwidth consumption, In this paper, a P2P VOD node selection strategy based on linear programming and reinforcement learning is proposed. The strategy is implemented in two layers: in the first layer, the server utilizes the global information collected periodically from the network node. Linear programming, In order to guide the node selection process, the second layer node selection strategy adopts cooperative Q-learning method, considering the local information of neighbor nodes, such as the remaining uplink bandwidth, the number of resources and the congestion state, etc. The experimental results show that the proposed strategy can effectively select the appropriate resource node set to provide resources and reduce the bandwidth consumption of the server effectively. Reduce server computation.
【学位授予单位】：江苏大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN948.64

【相似文献】