首页> 外文会议>Very large data bases >GHOST: Fine Granularity Buffering of Index
【24h】

GHOST: Fine Granularity Buffering of Index

机译:GHOST:索引的细粒度缓冲

获取原文
获取原文并翻译 | 示例

摘要

Buffering has all along been an important strategy for exploiting the cost/performance ratio of disk versus random-access memory. The buffering of disk pages belonging to a database has been well-studied, but literature that deals specifically with inder buffering is scarce. This is surprising given the significance of indexes (especially B+-tree like indexes ) in modern DBMSs. In this paper, we describe a dual buffering scheme for indexes, called GHOST, in which part of the buffer is used to maintain popularly used "paths" of the B+-tree index, while the remainder is devoted to maintaining a Splay-tree with pointers to leaf pages containing frequently used leaf pages. This scheme allows us to maintain pointers to leaf nodes long after the paths leading to the leaf nodes have been replaced, thus maintaining "ghost" paths to the nodes. In addition to describing the search and maintenance operations for the GHOST buffering scheme, we also conduct a series of experiments in which it is shown that GHOST outperforms the best existing schemes (ILRU and OLRU) by impressive margins for almost all pragmatic query workloads.
机译:一直以来,缓冲一直是利用磁盘与随机存取存储器的成本/性能比的重要策略。属于数据库的磁盘页面的缓冲已得到充分研究,但是很少有专门涉及内部缓冲的文献。考虑到索引在现代DBMS中的重要性(尤其是B +树式索引),这令人惊讶。在本文中,我们描述了一种称为GHOST的索引双重缓冲方案,其中部分缓冲区用于维护B + -tree索引的常用“路径”,而其余部分专门用于维护具有以下内容的Splay-tree:指向包含常用叶子页的叶子页的指针。这种方案使我们可以在指向叶子节点的路径被替换之后很长时间维护指向叶子节点的指针,从而维护到节点的“幽灵”路径。除了描述GHOST缓冲方案的搜索和维护操作之外,我们还进行了一系列实验,结果表明,对于几乎所有实用的查询工作负载,GHOST均以可观的优势胜过现有的最佳方案(ILRU和OLRU)。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号