Web缓存与预取模型研究

【作者】 卫琳；

【导师】 石磊；

【作者基本信息】 郑州大学 ， 软件工程， 2006， 硕士

【摘要】 随着互联网信息及用户的飞速增长，如何有效减少用户访问延时，提高网络服务质量是一个迫切需要解决的难题。缓存与预取技术是克服此难题的有效方法。本文针对Web对象访问特征、Web缓存与预取一体化模型等问题进行了研究。论文阐述了缓存和预取技术的基本概念、缓存系统和预取系统的分类与结构，以及缓存与预取模型的工作原理。根据用户访问特点，提出了预取技术框架。介绍了Web对象浏览特征，研究Web访问特征是有效进行Web缓存与预取的基础。针对Web访问特征，讨论了Web对象访问特征建模问题，提出了Web缓存与预取一体化模型(IWCPM)，并对IWCPM进行了性能测试和评价。本文的主要研究工作有以下几方面。1、分析Web对象访问特征，采用数学模拟方法分别对Web对象高频区及低频区流行度特征、Web对象大小重尾分布特征、Web访问的时间局部性等特征建模。设计并实现了一个Web特征模拟生成器WEBGEN。该模拟器可以模拟生成Web对象访问日志，具有较大的灵活性，为进一步研究Web缓存技术和预取技术提供依据。2、目前对于预取技术与缓存技术的研究往往强调预取算法或替换策略的改进，缺乏对缓存与预取一体化模型的研究。本文提出了Web缓存与预取一体化模型(IWCPM)并描述了其工作过程，分析和比较了仅缓存模型CM与IWCPM的性能指标，指出缓存与预取结合技术比仅缓存技术具有更好的性能表现。同时，结合Web访问特征，重点讨论了改进的PPM预测算法。3、设计实现了Web缓存与预取模拟器。比较了基于流行度的预测模型和基于访问模式的预测模型，实验中结合了四种典型的缓存替换策略GDSF，GDSize，LFU，LRU。缓存与预取一体化代表了今后提高Web性能的研究方向，性能分析给出了在缓存与预取一体化模型中，基于流行度的预测模型和基于访问模式的预测模型各自的适用范围。更多还原

【Abstract】 With the remarkable and exponential growth rate of Web information and users, how to reduce the user perceived access latency and improve the quality of service of the network is becoming a crucial problem, and Web prefetching and Web caching are the primary solutions. This paper has deeply studied the modeling of Web access characteristics, the integrated model for Web caching and prefetching.The concept, classification, structure of Web caching and prefetching and the working principle of integrated Web caching and prefetching system are described. A prefetch framework is proposed according to the user surfing procedure. Afterwards, Web surfing characteristics are discussed. Understanding the WWW traffic characteristics is the key to the effective design of Web caching and prefetching algorithm.Based on the Web access characteristics, the mathematical model of Web traffic is discussed, the integrated model of Web caching and prefetching is put forward. Then, the performance evaluation of the integrated Web caching and prefetching model (IWCPM) is made and discussed.The main research work of this thesis can be described as follows:1. This paper makes use of mathematical analytical approach to design and implement a Web LOG simulator: WEBGEN, in which the Web object popularity distribution, Web object size distribution and Web temporal locality are simulated. It not only can synthesize Web object access workload, but also has higher flexibility, and provide basis for further studying Web caching and prefetching.2. Previous studies in Web caching and prefetching mainly focus on improving replacement policies and building access models and evaluating the performance of such models in predicting future accesses. While these models are important, they lack the consideration of analysis based on integrated Web caching and prefetching system. Integrated Web caching and prefetching model (IWCPM) is presented and the performance evaluation of IWCPM is made. An improved PPM algorithm is discussed in detail in the discussion of the integrated model.3. A Web caching and prefetching simulator is designed and implemented. Experiments have been made based on IWCPM with the support of popularity-based prediction and pattern-based prediction mechanisms and four tipical replacement policies: GDSF, GDSize, LFU, and LRU. The experimental results illustrate the combination of Web prefetching and caching holds the promise of improving the QoS of Web systems. The corresponding application fields of popularity-based prediction and pattern-based prediction mechanisms are also presented in the paper.更多还原