Research On The Storage Of Condensed Cube Based On Flash Memory

The rapid response of a query OLAP multidimensional cube needs to pre-calculatedata and save the results. Due to the enormous size of the data cube, the response of a query is slowed down. Real data cube in memory, shortens the query response time by avoiding a large number of I /O operations, but the size of the data cube memory space is generally several orders of magnitude. Condensed Cube will condense gathered with a group of the cube of the basic unit of group into a set of tuples, while it effectively reduces the size of the data cube, but because of memorie's limitations,still can not satisfies the large memory data cube implementation of requirements.SSD (solid state drive) as a new type of secondary storage device, and its space and the speed of access can be a very good guarantee. Based on SSD features, combined with solid-state memory and the performance of hard drive two-level memory-SSD storage structure is proposed. Coarse granularity of cuboids is materialized in memory, In SSD implementation of fine-grained meta-group is materialized, which first of all ensure that all the finest granularity tuples are materialized in SSD. Memory for the implementation of cuboids method, in which the bottleneck re-calculation, consider the complete cubic stored in SSD. Point queries firstly response the cuboids materialized by memory, there is no cuboids queried in memory, then query in SSD. Range queries response directly in SSD. Two-level storage structure is used to make sure of the speed of point queries and the correctness of range queries.SSD as a flash, we must consider the characteristics of flash memory. Flash memory has asymmetric read and write in speed, it can not be repeated a limited number of write and erase characteristics. Transplanted directly to the SSD on the traditional index, certainly can not take full advantage of SSD's high-performance features. So for the SSD's I/O features and the current index structure based SSD problems, we proposed multi-level dynamic hash index structure characteristic of the hybrid ,thus make sure the efficient response of inquires.