Abstract: We present the interpolation search B-tree (ISB-tree), a new cache-aware indexingscheme
that supports update operations (insertions and deletions) in O(1) worst-case block
transfers and search operations in O(logB logn) expected block transfers, where B
represents the disk block size and n denotes the number of stored elements. The expected
search bound holds with high probability for a large class of (unknown) input distributions.
The worst-case search bound of our indexingscheme is O(logB n) block transfers. Our
update and expected search bounds constitute a considerable improvement over the
O(logB n) worst-case block transfer bounds for search and update operations achieved
by the B-tree and its numerous variants. This is also verified by an accompanying
experimental study.
Abstract: We present the interpolation search tree (ISB-tree), a new cache-aware indexingscheme that supports update operations (insertions and deletions) in O(1) worst-case (w.c.) block transfers and search operations in O(logB log n) expected block transfers, where B represents the disk block size and n denotes the number of stored elements. The expected search bound holds with high probability for a large class of (unknown) input distributions. The w.c. search bound of our indexingscheme is O(logB n) block transfers. Our update and expected search bounds constitute a considerable improvement over the O(logB n) w.c. block transfer bounds for search and update operations achieved by the B-tree and its numerous variants. This is also suggested by a set of preliminary experiments we have carried out. Our indexingscheme is based on an externalization of a main memory data structure based on interpolation search.
