在线多播编码缓存的研究

发布时间：2018-06-20 02:42

本文选题：编码缓存 + 中心化　；参考：《西南交通大学》2017年硕士论文

【摘要】：随着网络基础设施的发展和移动设备的普及,在线视频流量增长迅速,在网络流量中占比越来越大。在线视频具有时变性,高峰期带来的网络流量远大于低峰期网络流量,如果满足高峰期用户请求,会在低峰期造成很大的带宽浪费。编码缓存在低峰期将部分文件缓存在用户本地存储中,高峰期用户请求时,已经缓存的数据不需要从服务器发送,从而降低高峰期网络流量。此外,服务器利用用户间的缓存关系进行编码多播,在一个发送时隙内能够满足多个用户的请求,可以进一步减少高峰期网络流量。论文主要研究在线编码缓存问题。在线编码缓存在编码缓存基础上,增加了缓存的更新过程,以保证用户缓存的有效性。现有的在线编码缓存方案都是基于去中心化方案,而中心化方案相比去中心化方案有着更优的性能。论文将中心化方案应用于在线场景下,结果表明在活动用户数目不变时能获得性能增益。此外,论文对活动用户数目减少的情形也进行了分析,结果表明仍然能够满足用户请求,不过性能不是最优的。现有的在线编码缓存方案在更新服务器文件时,从服务器中随机选择一个文件进行替换,然后通过将用户缓存和服务器文件保持一致的方式进一步更新用户缓存。当服务器文件流行度非均匀分布时,随机替换的方式可能会导致替换出去流行度高的文件,从而降低缓存效率。论文使用最近最少发送的文件剔除规则优化现有方案,即用新文件替换服务器最近最少发送的文件,能够在最大程度上保证用户端缓存效率。结果表明优化后方案在非均匀分布下性能优于原来的方案。接着,在优化方案的基础上,进行在线分组,使得流行度高的组分配到更多的缓存空间,从而用户能够获得更好的缓存效率。结果表明分组方案在优化方案的基础上获得了有效的性能提升。
[Abstract]:With the development of network infrastructure and the popularity of mobile devices, the online video traffic is growing rapidly and accounts for more and more of the network traffic. Online video is time-varying, and the peak network traffic is much larger than the low peak period network traffic. If the peak users' requests are satisfied, it will cause a great waste of bandwidth in the low peak period. The encoded cache caches some files in the local storage of the user during the low peak period. When the user requests during the peak period, the cached data does not need to be sent from the server, thus reducing the peak network traffic. In addition, the server encodes multicast using the cache relationship between users, which can satisfy the requests of multiple users in one transmission slot, and further reduce the peak network traffic. In this paper, the problem of online coding cache is studied. On the basis of encoding cache, the update process of cache is added to ensure the effectiveness of user cache. The existing online coding cache schemes are based on the de-centralization scheme, and the centralization scheme has better performance than the de-centralization scheme. In this paper, the centralization scheme is applied to the online scenario. The results show that the performance gain can be achieved when the number of active users is constant. In addition, the reduction of the number of active users is also analyzed. The results show that it can still satisfy the user requests, but the performance is not optimal. The existing online encoding cache scheme selects a file randomly from the server to replace it when updating the server file, and then updates the user cache by keeping the user cache consistent with the server file. When the server file popularity is not uniformly distributed, the random replacement may lead to the replacement of high popularity files, thus reducing the cache efficiency. This paper optimizes the existing scheme by using the least recently sent file removal rule, that is, replacing the most recently sent files with new files, which can guarantee the efficiency of client cache to the greatest extent. The results show that the performance of the optimized scheme is better than that of the original scheme under non-uniform distribution. Then, on the basis of the optimization scheme, the group with high popularity can be allocated more cache space, so that users can obtain better cache efficiency. The results show that the performance of the grouping scheme is improved effectively on the basis of the optimized scheme.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN919.8

【相似文献】