The GiST project studies the engineering and mathematics behind
content-based indexing for massive amounts of complex content. The
project consists of a number of components:
- The basis of our work is the Generalized
Search Tree (GiST), a template indexing structure that allows
domain experts (e.g in computer vision, bioinformatics, or remote
sensing) to easily customize a database system to index their content.
- The GiST allows any workload to be indexed, but not every workload
can be indexed efficiently. We are mapping the boundaries of
indexing efficiency by developing a Theory of Indexability to describe
how hard an indexing problem is.
- Even for indexable problems, the task of developing an efficient
index can be tricky. Amdb is a tool we
developed that allows index extenders to tune and debug their
index extensions. Amdb combined debugging and visualization features
with an analysis framework that allows developers to compare their
performance to the theoretical optimum.
gist@db.cs.berkeley.edu
Last modified: $Date: 1999/08/17 02:13:24 $ by $Author: jmh $