【作者】 许志闻；

【导师】 庞云阶；

【作者基本信息】 吉林大学 ， 计算机应用技术， 2004， 博士

【摘要】 今天，我们已经进入了数字化的信息时代，随着Internet技术的发展和普遍应用，网络中50%以上的存储内容是流媒体数据；流媒体的应用系统、国际标准和基础研究正成为目前产业界和科研机构研究关注的焦点，解决流媒体广泛应用和网络资源紧张的矛盾是具有挑战性的研究课题，本文开展了流媒体代理缓存高效技术的研究，提出了动态缓存算法，设计了三种动态缓存算法，并将动态缓存算法应用在基于前缀缓存、分段缓存和视频组播技术中；对各种传送方案中的调度策略进行改进和提高；建立数学模型，定量分析动态缓存算法高效性能；并评估了高效动态缓存算法、分段缓存算法、前缀缓存算法和全部缓存策略的性能，进一步说明了动态缓存的高效性能。动态缓存和分段缓存都能提高流媒体代理的高效性能，文中给出了分段缓存和动态缓存的分配关系，给出了具有最佳效果的分段动态缓存算法，将分段动态缓存算法应用于补丁算法，建立基于动态缓存的成批补丁算法，其具有较高的效率。本文的具体贡献和研究结果如下：1． 对流媒体代理的缓存调度技术进行了全面的综述。从分析讨论基于Internet的代理缓存技术出发，全面介绍了流媒体调度的各种算法，重点介绍了流媒体传送过程中的流调度中取得的研究成果，说明了补丁算法、成批算法的研究现状、应用领域、优势与不足，阐述了研究这些问题的数学模型。这些内容的讨论和分析是开展进一步研究的基础。2． 研究了流媒体代理缓存中的基于前缀的高效技术。在分析代理缓存的前缀缓存技术基础上，提出了动态缓存的三种算法，基于窗口、数量和高效算法；将三种算法分别应用于单播后缀成批、单播补丁、多播补丁、多播流合并和成批补丁等传送方案中，都能较好地提高代理缓存的效率。根据客户请求过程符合泊松分布，建立了各种传送方案的数学分析模型，定量地分析了动态缓存算法在各种传送方案下，对流媒体系统性能的提高。3． 研究了基于分段的动态缓存算法。分段缓存方法是一种有效的高效缓存方法，在分段缓存中缓存的大小根据客户的请求频率进行分配，分段缓存和前缀缓存都缓存在代理缓存中；基于分段的动态缓存在分段缓存的基础上，进一步提高代理缓存的效率，将三种<WP=157>基于窗口、数量和高效算法的动态缓存算法分别应用在基于分段的单播后缀成批、单播补丁、多播补丁、多播流合并和成批补丁等传送方案中，方便、高效地提高了代理缓存的效率；通过建立各种传送方案的数学分析模型，定量地分析了动态缓存算法在基于分段各种传送方案下，对流媒体系统性能的提高。分析采用客户端缓存将动态缓存分成动态缓存窗和动态缓存两种应用方式，进一步提高代理缓存的命中率。分段缓存和动态缓存都能较好地提高代理缓存的效率，给出了两者的分配关系，定量地说明了两者对代理缓存效率提高的程度；动态缓存不需要长时间保存在代理缓存中，分段缓存需要较长时间内保存在代理缓存中，分段缓存提高效率是通过早期的客户请求进行判断，而动态缓存可以通过当时的客户请求行为进行处理，分段缓存和动态缓存算法同时使用，可以更好地提高代理缓存的效率。评估了基于分段的高效动态缓存、分段缓存、前缀缓存和全部缓存的性能，结果高效动态缓存性能最优。5． 研究了动态缓存在组播技术中的应用。组播技术是节约网络资源的有效方法，为了解决流媒体应用于网络资源不足之间的矛盾，视频组播技术成为网络流媒体研究的一个重要领域，视频组播产生客户的等待延迟，同时要求较高的使用条件，基于代理缓存的应用级组播是解决视频组播这一问题的有效方法。流媒体代理缓存通过使用动态缓存技术，可以有效地进行组播流的合并、流媒体的静态和动态算法的调度；使用动态缓存的静态调度算法，拓宽成批传送算法的长度，而且不会产生客户的等待延迟；动态缓存可以替代周期广播算法使用。使用动态缓存替代周期性补丁算法、选择性周期补丁算法，节约补丁算法占用的网络和服务器资源；使用动态缓存可以改进补丁优先算法，拓宽补丁优先算法的性能和应用范围，提高各种算法的效率和应用范围。6． 介绍了代理缓存使用动态缓存算法的设计中的一些关键问题。介绍了前缀缓存、分段缓存和动态缓存的时序安排和工作过程；根据视频服务器设计中取得的研究成果，介绍了流媒体代理缓存的接入控制和替换策略；采用视频服务器中使用的间隔缓存和分组缓存策略，说明流媒体代理缓存的磁盘替换方法。本文的对流媒体代理缓存高效技术进行的研究，提出了动态缓存的高效方法，应用动态缓存算法较大幅度地提高了多种流媒体调度算法的性能，流媒体代理的动态缓存算法是一种高效、实用的缓存算法，它的研究具有一定的理论意义和应用价值。更多还原

【Abstract】 In the age of digital information that we have entered, with the rapid development and widespread application of Internet technology, more than 50% of Internet data is stored in the formal for streaming media, whose application system, international standard and basic research have become one of the focuses of the domains for current industry and science research. To address the conflict between wide application of streaming media and limited Internet’s resources are one of challenging projects. This thesis carries out some research on the proxy cache for streaming media; proposes algorithm for dynamic ache and applies it to the prefix cache, segment-based cache and video multicast; improves the caching policy in every transmitting project; sets up mathematic model; presents quantitative analysis of algorithm for dynamic cache and evaluates efficiency of the algorithms of dynamic cache, segment-based cache, prefix cache and full cache and finally gives a further description of the high-powered dynamic cache. Both dynamic cache and segment-based cache can improve high-powered proxy cache for streaming media. Thus this thesis presents the relationship of the distribution between the segmented caching and dynamic caching algorithm. This thesis applies the algorithm of dynamic cache to patching and batching, and further, present the algorithm of batched patching using dynamic cache is optimal effective.The main contributions and results included in thesis are as follow:Firstly, a detailed of the literature review of the transmitting policy of proxy cache for streaming media is described. This literature review begins with the Internet-based proxy cache, and then introduces all kinds of algorithms of the transmitting policy of streaming media and highlights the research results of streaming scheduling policy in the transmitting process of streaming media, and illustrates the current situation, the applied field, the mathematic model, and advantage and disadvantage of the patching algorithm and the batching algorithm. All above this technique serve as the basis of further research.Secondly, the thesis researches on the high-powered technique of prefix cache for streaming media. This section, on the basis of prefix cache, puts forward three kinds of algorithms of <WP=159>dynamic cache: window-based, quantity-based and effect-based algorithm, which are proved to be able to improve the efficiency of proxy cache, when applied respectively to unicast suffix batching, unicast patching, multicast patching, streaming merging, batched patching and other transmitting projects. In accordance with clients’ request, it sets up mathematic models for various transmitting projects and presents quantitative analysis of how dynamic cache improves the efficiency of steaming media in every transmitting project.Thirdly, the thesis researches on the algorithms of dynamic cache based on segmented cache. Segmented cache is one of the high-powered caching methods, the caching size of which is allocated in accordance with request frequency of clients, and both segmented cache and prefix cache are cached in the proxy cache. Based on the segment-based cache the dynamic cache further enhances the efficiency of the proxy cache, by applying the three dynamic caching algorithms of window-based, quantity-based and effect-based to segment-based unicast suffix batching, unicast patching, multicast patching, streaming media and batched patching and other transmitting projects. This section also presents quantitative analysis of how dynamic caching algorithm improves the efficiency of streaming media system in various segment-based transmitting projects, by setting up the mathematic analysis model for various transmitting projects. It also analyzes how dynamic cache, which can be divided into the following two kinds of application forms: dynamic caching window and dynamic cache improves the caching efficiency. Both segment-based cache and dynamic cache can improve the efficiency of proxy cache, and their distributive relationship is illustrated更多还原