基于关联性特征的视频点播关键问题优化研究
发布时间：2018-09-18 11:26
【摘要】：视频点播VoD（Video-on-Demand）实现了用户与视频资源高度交互的访问模式，，使其得到越来越多用户的青睐，成为目前互联网上需求量最大的服务之一。点播服务中用户交互式操作、用户实时性观看需求、服务规模爆炸性增长及用户视频数据请求量不稳定性等特点给实现一种高用户满意度且经济有效的点播服务带来了一定的问题。考虑到点播系统中视频片段及视频间存在的关联性特征，本文基于关联性特征对如何实现经济有效且高质量的点播服务进行了优化研究，主要从以下几个问题进行展开： 1）快速资源定位机制的实现。针对该问题，提出一种基于视频片段关联性的点播服务模型。所提模型利用视频点播中视频各个片段之间关联性强度不同这一特点，使具有相似兴趣点的节点能够自主地组织在一起，从而有效地提高了节点在进行转跳操作时查找目的片段的效率，提升了用户观看满意度。对由该策略引出的节点邻居优化问题进行了分析，将其形式化描述为集合覆盖问题，并提出一种基于贪心算法的邻居近似优化策略来解决该问题。仿真对比实验验证了所提策略在降低查找时延方面的有效性，并对所提邻居优化算法的有效性进行了验证。 2）有关联性视频的点播服务在云平台上的部署。针对该问题，本文通过对YouTube上有关联性关系视频的采集，得出用户以近似90%的概率会选择当前视频三跳关联以内的视频进行观看这一结论，基于此提出一种基于簇的P2P云辅助点播服务模型。在所提模型中，由于云服务器存在一定的启动延迟，因此对簇的带宽迁移量进行了预测；针对云服务器有着不同的类型、租用代价以及服务能力的特点，提出一种满足用户带宽需求前提下降低云服务器租用代价的云服务器租用策略，仿真对比实验验证了所提策略在降低服务资源开销以及提高用户满意度上的有效性。 3）关联性视频内容在分布式云服务器上的内容放置问题。针对该问题，考虑到关联性视频的区域流行度差异，提出一种地域感知的关联性视频内容在分布式云服务器上的内容放置策略。该策略目的是在尽量维持关联性关系及不同云服务器上负载均衡的同时降低用户跨地域请求率。将问题形式化描述为一个有限制的K中心点划分问题，提出一种地域感知的PAM算法解决该问题。仿真对比实验表明了该算法在提高用户本地请求率和降低用户跨区域请求方面的有效性，同时也能较好地维持视频之间的关联关系。 4）提高节点缓存空间利用率的关联性视频内容缓存替换问题。针对此问题，提出一种基于视频相似的缓存替换策略。所提策略优先考虑替换同已替换视频集合语义相似度较大即可能为不受欢迎的视频，并使替换视频整体流行度尽可能地小且副本数尽可能地大。该问题可描述为一个多目标优化问题，将其转换为单目标优化问题并进一步给出缓存替换策略。仿真实验分析了策略中参数设置对缓存命中率的影响，并证实了所提策略在提高缓存内容命中率上的有效性。 本论文研究得到国家“863计划”专项课题（2008AA01A315），教育部高等学校博士学科点专项科研基金课题“基于云服务的视频点播关键技术研究”(20114101110007)，河南省科技创新人才计划项目“P2P VoD关键技术问题研究”（2011HASTIT003)以及河南省教育厅重点项目“基于云平台的P2P VoD关键技术研究”(13A520562)的资助。
[Abstract]:Video-on-Demand (VOD) has become one of the most demanding services on the Internet because of its highly interactive access mode between users and video resources. Considering the correlation between video clips and videos in VOD system, this paper studies how to optimize the cost-effective and high-quality VOD service based on the correlation characteristics. We should start with the following issues:
1) Realization of a fast resource location mechanism. To solve this problem, an on-demand service model based on video clip association is proposed. The proposed model takes advantage of the different correlation strength between video clips in VOD, so that nodes with similar interest points can be organized independently, thus effectively improving the nodes. The efficiency of searching for the target segment during the jump operation improves the user's viewing satisfaction. The neighborhood optimization problem derived from this strategy is analyzed and formally described as a set coverage problem. A neighborhood approximate optimization strategy based on greedy algorithm is proposed to solve the problem. The proposed strategy is effective in reducing the search delay, and the validity of the proposed neighborhood optimization algorithm is verified.
2) Deployment of associative video on the cloud platform. To solve this problem, this paper collects related video on YouTube, and concludes that users will choose the video within the current three-hop association with a probability of 90%. Based on this conclusion, a cluster-based P2P cloud-assisted VOD service is proposed. In the proposed model, the bandwidth migration of the cluster is predicted because of the delay of the start-up of the cloud server, and a rental policy of the cloud server is proposed to reduce the rental cost of the cloud server while satisfying the user's bandwidth requirement. Simply, the simulation results verify the effectiveness of the proposed strategy in reducing service resource overhead and improving user satisfaction.
3) Content placement of related video on distributed cloud servers. To solve this problem, considering the regional popularity difference of related video, a region-aware content placement strategy for related video on distributed cloud servers is proposed. The problem is formally described as a restricted K-center partitioning problem, and a region-aware PAM algorithm is proposed to solve this problem. Simulation results show that the algorithm is effective in improving the local request rate and reducing the cross-region request rate. It can also better maintain the correlation between videos.
4) Relevant video content cache replacement problem to improve node cache space utilization. To solve this problem, a video-based cache replacement strategy is proposed. The proposed strategy gives priority to replacing videos that may be undesirable if they have a large semantic similarity with the replaced video set, and makes the replacement video as popular as possible. This problem can be described as a multi-objective optimization problem, which can be transformed into a single-objective optimization problem and a cache replacement strategy is given. The simulation results show that the proposed strategy is effective in improving the cache hit rate.
In this paper, the National 863 Program (2008 AA01A315), the Ministry of Education's Ph.D. Program Research Foundation Project "Research on Key Technologies of VOD Based on Cloud Services" (20114101110007), the Henan Science and Technology Innovation Talents Program Project "Research on Key Technologies of P2P VoD" (2011 HASTIT003) and The key project of Henan Provincial Department of education is supported by "P2P VoD key technology research based on cloud platform" (13A520562).
【学位授予单位】：郑州大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN948.64
本文编号：2247759
[Abstract]:Video-on-Demand (VOD) has become one of the most demanding services on the Internet because of its highly interactive access mode between users and video resources. Considering the correlation between video clips and videos in VOD system, this paper studies how to optimize the cost-effective and high-quality VOD service based on the correlation characteristics. We should start with the following issues:
1) Realization of a fast resource location mechanism. To solve this problem, an on-demand service model based on video clip association is proposed. The proposed model takes advantage of the different correlation strength between video clips in VOD, so that nodes with similar interest points can be organized independently, thus effectively improving the nodes. The efficiency of searching for the target segment during the jump operation improves the user's viewing satisfaction. The neighborhood optimization problem derived from this strategy is analyzed and formally described as a set coverage problem. A neighborhood approximate optimization strategy based on greedy algorithm is proposed to solve the problem. The proposed strategy is effective in reducing the search delay, and the validity of the proposed neighborhood optimization algorithm is verified.
2) Deployment of associative video on the cloud platform. To solve this problem, this paper collects related video on YouTube, and concludes that users will choose the video within the current three-hop association with a probability of 90%. Based on this conclusion, a cluster-based P2P cloud-assisted VOD service is proposed. In the proposed model, the bandwidth migration of the cluster is predicted because of the delay of the start-up of the cloud server, and a rental policy of the cloud server is proposed to reduce the rental cost of the cloud server while satisfying the user's bandwidth requirement. Simply, the simulation results verify the effectiveness of the proposed strategy in reducing service resource overhead and improving user satisfaction.
3) Content placement of related video on distributed cloud servers. To solve this problem, considering the regional popularity difference of related video, a region-aware content placement strategy for related video on distributed cloud servers is proposed. The problem is formally described as a restricted K-center partitioning problem, and a region-aware PAM algorithm is proposed to solve this problem. Simulation results show that the algorithm is effective in improving the local request rate and reducing the cross-region request rate. It can also better maintain the correlation between videos.
4) Relevant video content cache replacement problem to improve node cache space utilization. To solve this problem, a video-based cache replacement strategy is proposed. The proposed strategy gives priority to replacing videos that may be undesirable if they have a large semantic similarity with the replaced video set, and makes the replacement video as popular as possible. This problem can be described as a multi-objective optimization problem, which can be transformed into a single-objective optimization problem and a cache replacement strategy is given. The simulation results show that the proposed strategy is effective in improving the cache hit rate.
In this paper, the National 863 Program (2008 AA01A315), the Ministry of Education's Ph.D. Program Research Foundation Project "Research on Key Technologies of VOD Based on Cloud Services" (20114101110007), the Henan Science and Technology Innovation Talents Program Project "Research on Key Technologies of P2P VoD" (2011 HASTIT003) and The key project of Henan Provincial Department of education is supported by "P2P VoD key technology research based on cloud platform" (13A520562).
【学位授予单位】：郑州大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN948.64
【参考文献】
相关期刊论文 前4条
1 吴松;金海;邹德清;;一种流媒体文件的分块放置方法[J];计算机学报;2006年03期
2 刘亚杰;窦文华;;一种P2P环境下的VoD流媒体服务体系[J];软件学报;2006年04期
3 胡懋智;徐恪;夏树涛;叶明江;;TOW:一种新的P2P实时流媒体缓存替换算法[J];小型微型计算机系统;2009年08期
4 王娟;纪其进;朱艳琴;;基于用户行为特征的P2P视频点播系统数据预取机制[J];小型微型计算机系统;2010年10期
本文编号：2247759
本文链接：https://www.wllwen.com/kejilunwen/wltx/2247759.html
