Improving Performance with Bulk-Inserts in Oracle R-Trees

摘要

Spatial indexes play a major role in fast access to spatial and location data. Most commercial applications insert new data in bulk: in batches or arrays. In this paper, we propose a novel bulk insertion technique for R-Trees that is fast and does not compromise on the quality of the resulting index. We present our experiences with incorporating the proposed bulk insertion strategies into Oracle 10i. Experiments with real datasets show that our bulk insertion strategy improves performance of insert operations by 50%-90%.