The spatial c-tree: an index structure with high concurrency and efficient recovery

Indexes are used in database management systems as a mechanism to speed up data retrieval. A good index structure should ensure that the data access through it is efficient and transactional. For one dimensional data, the B+-tree and its variants, such as the B link-Tree and the Π-tree, are used widely because of their good query performance and concurrency control and recovery. However, structures for multidimensional data, despite the intense research for more than two decades, have not been so successful. Very few of the many proposed structures are implemented in commercial databases. For those that have been used, users are confronted with either poor performance or low concurrency.We proposed a new spatial access method called the Spatial C-Tree. Our approach extends the C-Tree framework, which provides efficient concurrency control and robust recovery mechanism. We have implemented the Spatial C-Tree and compared it with several R-Tree variants. Our performance results show that the Spatial C-Tree performs very well in terms of insertion and exact match queries with various data sets. Our range query performance in some cases is comparable to the R-Tree variants and in other cases are outperformed by the R-Tree variants. We have identified possible improvements over the range query performance. The potential improvement on the range query performance and the high concurrency guarantee make the Spatial C-Tree a promising access method for spatial data.