多维空间数据索引结构的并发控制方案

Concurrency Control Scheme of Index Structure for Multi-Dimensional Spatial Data

R树是支持多维空间数据访问的重要索引结构之一 ,但由于缺乏适用的并发控制机制来保证并发环境下数据的一致性 ,因此至今大多数商用数据库系统都不支持基于 R树的并发处理。建立在 R树基础上的 R- link树解决了并发控制中出现的部分难题 ,但它仍然存在幻像等问题 ,因此分析 R- link树中尚存的问题 ,并通过设计一个基于内存的操作控制列表来预先避免可能冲突的并发操作 ,从而实现完全的并发控制。

R tree is an important index structure for spatial data access. But most commercial spatial database systems do not support R tree based concurrent processing due to the lack of protocol to provide consistency in concurrent environment. R link tree is a variation of R-tree and it can deal with this issue to some extent, but such problems as phantom still exist. In this paper, the author analyzes these problems and suggests an in memory operation control list (OCList) to prevent conflicting operations and thus realize the complete concurrency control. The experiment results show that this scheme is correct and it can improve the system performance.