Module 5: Disk-based Index Structures
  Lecture 17: Disk-based Index Structures
 

                                           

 

 

K-d-B tree
  • Bucket version of K-d trees
  • Combine multi-dimensional search effciency of K-d trees with disk I/O effciency of B-tree
  • Balanced
  • Point pages: Leaf nodes
 
  • All data are in the leaf nodes
  • Region pages: Internal nodes
 
  • Contains region keys
 
  • Region splits (hyperplanes) need not alternate
  • Under flow restriction of B-trees is relaxed (B'-tree)
 
  • Space utilization is low