Create a Rtree using a Map Reduce algorithm?

Question

In the current scenario, I Have a Rtree instance in which I add millions of records, which takes around 1 hr to create. I was wondering if I can use multiple mappers to create multiple RTrees & then merge them in the reducer to create the final RTree? Is there a specific merging Rtree technique available? How should I go about solving this? Any help is well appreciated?

Answer 1

Have you considered bulk-loading the index? Common bulk-loading strategies are based on sorting, which is well understood how to do in a distributed, scalable manner. This way, I've build R-trees with millions of records in just a few seconds.