基于频繁序列挖掘的预取算法研究与实现

发布时间：2018-05-23 15:15

本文选题：频繁序列挖掘 + 预取算法　；参考：《计算机研究与发展》2016年02期

【摘要】：预取作为一种提升存储系统性能的有效手段被广泛使用,然而传统的预取算法大多基于顺序性访问特征的探测,这使得它们在非顺序数据访问环境下很难奏效,甚至可能因为预取准确率较低而对存储系统的性能带来负面影响.而基于频繁序列挖掘的预取算法则能够通过分析数据的访问行为找出潜在规律,从而能在非顺序访问模式下也取得一定的性能提升.同时,为了应对某些缓存受限的应用场景,如嵌入式系统,预取算法通过提高分析的准确率减少预取可能对缓存带来的不利影响.新提出的预取算法基于频繁序列挖掘技术,并使用字典树组织预取规则,通过多步匹配和子树分割技术精细地控制规则的使用,提升预取的准确率,从而使得预取算法能够有效提升存储系统的性能.
[Abstract]:Prefetching is widely used as an effective means to improve the performance of storage systems. However, traditional prefetching algorithms are mostly based on sequential access feature detection, which makes it difficult for them to work in non-sequential data access environments. It may even have a negative impact on storage system performance because of low prefetching accuracy. The prefetching algorithm based on frequent sequence mining can find out the potential rules by analyzing the access behavior of data, which can also improve the performance in non-sequential access mode. At the same time, in order to deal with some application scenarios with limited cache, such as embedded system, prefetching algorithm can reduce the adverse effect of prefetching on cache by improving the accuracy of analysis. The proposed prefetching algorithm is based on frequent sequence mining technology, and uses dictionary tree to organize prefetching rules. The rules are carefully controlled by multi-step matching and sub-tree segmentation techniques to improve the accuracy of prefetching. Thus the prefetching algorithm can effectively improve the performance of the storage system.
【作者单位】：武汉光电国家实验室(华中科技大学);华中科技大学计算机科学与技术学院;
【基金】：国家“八六三”高技术研究发展计划基金项目(2013AA013203) 华中科技大学自主创新研究基金项目(2014QN010)~~
【分类号】：TP311.13

【相似文献】