A self-adaptive improved μ-Tree index structure for flash-based DBMS

As a pure electronic device the flash memory has many outstanding features, such as higher read-write speed, lower noise and better shock-resistance. Flash-based DBMS soon becomes the area that attracts the most attention. A Flash-based DBMS FDBMS1.0 is implemented by us. The index is very important to improve the efficiency of flash-based DBMS, and researchers have put forward many index algorithm about it, among them, μ-Tree is a higher efficiency index. The μ-Tree is improvement of B+-tree for flash memory. One of great strengths of μ-Tree is the ability to remove the phenomenon of “wandering trees” that bring by traditional B+ Tree in flash memory. There is a scent of trouble to deal with update of index and the maintenance of tree's structure for intensive update or small record update operations. This paper proposes an improved μ-Tree structure to reduce cost in update index and the maintenance of tree's structure. The update-area and self-adaptive mechanism is utilized in the improved μ-Tree structure. The experimental results show that our method is effective.