摘要
目前基于局部匹配预测(PPM)模型的研究关注的焦点是在保证预测精度的前提下,尽量缩减PPM的空间占用,但缺乏自适应动态更新机制,难以实现在线预取。针对Web访问特点,提出了基于流行度的自适应预测模型。该模型的核心是基于Web对象流行度的PAPPM预取算法,通过模型构造、模型预测和模型更新三个过程实现了动态自适应的Web预取。讨论并实现了确定性上下文预测,最优阶估算以及上下文LRU替换策略等功能。在Web缓存与预取一体化条件下的实验表明,该模型具有较高的性能,适用于在线预取。
The current research of Prediction by Partial Match （PPM） model generally focuses on the reduction of space complexity of the model under the condition of guaranteeing the prediction accuracy. But most of the studies lack the adaptive mechanism, which is requisite in on-line systems. In terms of Web access characteristics, popularity based adaptive PPM prediction model （PA PPM） was proposed, whose core was prefetching algorithm based on Web objects＇ popularity, PA PPM actualized dynamic adaptive Web prefetching by three parts： model construction, model prediction and model update, The mechanisms of deterministic context prediction, optimal order estimation and LRU based discarding policy to support adaptation were discussed and realized. Under the condition of integrated Web caching and prefetching, experimental results have shown that PA PPM model can achieve a good performance and can be used to realize on-line Web prefetching.
出处
《计算机应用》
CSCD
北大核心
2008年第3期553-557,共5页
journal of Computer Applications
基金
国家自然科学基金资助项目(60472044)
河南省信息网络重点实验室开放基金项目资助项目(2006)
关键词
预取
局部匹配预测
自适应
最优阶估算
prefetching
Prediction by Partial Match （PPM）
adaptive
optimal order estimation
