Speeding up on-the-fly integration of DB and exo-DB data

Applications that perform on-the-fly integration of data inside a database (DB) with "exo-DB data " (that is, data outside a DB) are becoming common. Ease of integration and speed of performance are key considerations in such applications. To this end, we have devised a means to mix data in a traditional DB with references to fine-grained exo-DB data of arbitrary formats, and a means to use "bi-level queries " to integrate DB data and exo-DB data using existing query languages. In this paper, we present a "bulk accessor" to speed up retrieval of data from a large number of exo-DB references. The bulk accessor supports different policies a query processor can use to improve performance depending on data characteristics such as clustering. Experiments show that the bulk accessor saves 69%-90% time per query over a naive method.