Index method for main storage database based on T-lt tree

Abstract

The invention provides an index method for a main storage database based on a T-lt tree, improving the defect of the T tree, adding a tail pointer on each node of the T tree which directly points to a successor of the node to form a T-lt tree. The index method comprises a database inquiring operation, a keyword insertion operation and a keyword deleting operation. The node structure of the T-lt tree has one more tail pointer than the T node. If the node has a tail node, the tail pointer points to the tail node. If the node has no tail node, the tail pointer directly points to the successor. The tail node of the T-lt tree has one more pointer which directly points to the successor than the T node. The method reduces the traversal number to the tree, decrease the rotation number of the tree, can improve the inquiring efficiency in the main storage database and saves the space time spending.

Description

technical field [0001] The invention relates to a database indexing method, which is mainly applied in a database management system based on main storage, and belongs to the field of computer databases. Background technique [0002] As large-capacity memory becomes cheaper and cheaper, it is now easier to configure a computer with a large amount of memory. As far as the database system is concerned, it is already possible to keep the "main version" of the database in memory, which can greatly improve system performance, such as a large reduction in I / O operations, transaction state transitions and associated CPU caches Substitutions are greatly reduced, lock contention is reduced, more efficient in-memory lookup structures and query processing can be used, etc. [0003] The "master version" of the database or the database system in which the entire database is resident in memory is called a main memory database (MMDB, Main Memory Database). Due to the different characteris...

AI Technical Summary

Problems solved by technology

[0016] Among them, patent ② and papers ① and ② are all about the indexing mechanism of B-tree and B+ tree, which are suitable for disk-based database indexing mechanisms, but not for main memory databases, and are similar to balanced binary trees in terms of time and space overhead; patent ① and papers ③It is all about the indexing mechanism of T-tree. Although T-tree is an excellent indexing mechanism for databases based on main memory, updating the nodes of T-tree will cause the tree to be unbalanced. Therefore, the index used to balance rotation The cost is still relatively large

Images