An incremental batch-oriented index for bitemporal databases

Bitemporal databases record not only the history of tuples in temporal tables, but also record the history of the databases themselves. We address the problem of indexing such bitemporal databases by investigating the use of an incremental indexing structure, the HR-tree, which was originally aimed at spatiotemporal databases. The HR-tree's most attractive feature is that it can process queries as if all previous database snapshots were indexed physically, however, all such states are indexed only logically. In our experiments we have found that the HR-tree is much more efficient (up to 80% faster) than previously proposed approaches based on two coordinated R-trees when processing queries based on a single transaction time point and valid time being either point or intervals. As for size, the HR-tree was found to be better suited for workloads where the number of updates per transaction timestamp is reasonably large (over one thousand updates in our studies), otherwise it is prone to require large storage space.