A proposal for indexing bitemporal databases via cooperative B+-trees

We propose an indexing structure for bitemporal databases. Such structure is based on two trees, one indexing valid time and another indexing transaction time. The trees share pointers to the actual data records, which are thus not duplicated. Bitemporal queries are processed by dividing the query in two parts, a valid time part and a transaction time part. Each tree is searched according to these partial queries, and the answer is determined by the correct composition of the partial answers. We show how simple coordination of the tree searching, along with a simple assumption on the temporal data, improves query processing performance. The proposed structure also allows querying either time dimension, separately from the other one