【作者】 池静；

【导师】 牛保宁；

【作者基本信息】 太原理工大学 ， 计算机应用技术， 2003， 硕士

【摘要】 随着Internet技术和WWW服务的发展，Web网络流量的增加和网页访问的延迟日益引起人们的关注，这两个问题影响了Internet的持续发展。网络缓存技术是解决这两个问题的一种至关重要的技术，在国际上已经形成一个独立的主流研究领域，并取得了一些研究成果。网络缓存技术是一个复杂的课题，它需要解决替换策略、一致性维护、缓存共享和性能评价等诸多问题。虽然目前已经在这些方面做了很多工作，但许多问题并没有得到圆满解决，影响了网络缓存技术在WWW服务上的应用。本文的内容属于缓存共享领域。利用Bloom filter表示共享信息的内容，大大地降低了用于存储索引的空间消耗，减少了访问延迟。Bloom filter是一个简明的空间效率极高的随机的数据结构，用于判别一个元素是否属于某个集合。用Bloom filter表示cache内容，可以高效地实现cache协作。因为在代理之间只需传输Bloom filter而不是完整的cache目录表。 本文首先介绍了Bloom filter的研究和应用现状，然后，从数学角度对Bloom filter和Weighted Bloom filter进行比较。结果证明Weighted Bloom filter有较低的错误预测。但是，模拟结果显示，Bloom filter有较低的错误预测，比Weighted Bloom filter好。主要原因是Weighted Bloom filter需要很强的条件，而这些条件在现实中不能被满足。 太原理工大学工学硕士学位论文 本文最后指出Bloom ilter应用中存在问题和进一步研究的方向和措 施。更多还原

【Abstract】 With the development of Internet technologies and WWW services, more and more attention is paid to the Web traffic and page access delay , which are the key issues affecting the continuous growing of Internet . Proxy Cache,an effective technique to solve these issues, is a hot research area . Some research works have been conducted on this area and result in some fruits . Research on Proxy Cache is quiet complicated because there are a lot of problems such as replacement strategy, cache consistency, cache sharing and performance analysis. Though studying of Proxy Cache has been made of years, few of these problems have been solved successfully. This thesis, concentrates on topic of cache. The contents of shared message are represented by Bloom filter. This technique greatly reduces the space needed for page indexing and decrease the access delay. A Bloom filter is a simple space-efficient randomized data structure for representing a set in order to determine a certain element in the set . Weighted Bloom filters and counting Bloom filters have been suggested as means for sharing Web cache information . Bloom filters are transmitted among shared proxies instead of sending the full list of cache contents .In this thesis, a summary about the current research and application onBloom filter is presented, and then a mathematical comparison between the Bloom filter and weighted Bloom filter is given. It is proved that weighted Bloom filter has lower false prediction rate than Bloom filter. But the simulation results shown that Bloom filter has lower prediction rate and it is better than weighted Bloom Filter. The major reason is weighted Bloom filter needs very strong conditions, which cannot be satisfied in the real world.Finally, the problems in Bloom filter and the future research direction and measure are pointed out in this paper.更多还原