Barrodale Computing Services Announces GiSTXten for SQL Server 2008 Spatial Enhancements

"With the availability of 2D spatial indexing in SQL Server 2008, users will be able to build spatial indexes and perform efficient queries on their 2D spatial data," said Dr. Ian Barrodale, President of BCS. "A natural extension to these new spatial capabilities in SQL Server 2008 would be the ability to use powerful GiST indexing methods. These techniques are advantageous in database searching and scientific computing applications involving multidimensional or other complex data. GiST generalizes B-trees and R-trees, and is extensible both in the data types supported and in the queries that can be applied to the data. A particularly appealing feature of GiST is that it allows new data types to be indexed in a manner that supports queries which are natural to the data type."

BCS's GiSTXten allows users to develop GiST indexes tailored to their individual needs. It is implemented as a C# class with a set of methods for GiST index development. This class can then be subclassed by users to develop their own code implementing the required GiST index, which can then be loaded into C# and set up as an extension to SQL Server 2008. The use of GiSTXten for GiST index development helps simplify the implementation of these complex indexes and can provide an avenue for powerful and flexible database searches using these indexes.

One area where GiST indexing has been used is in applications involving multiple dimensions. An example taken from Life Sciences is to rapidly identify those chemical compounds with similar structural features to those of a target structure. This is often done using "fingerprints", which are bitmaps whose patterns are set based on preselected structural characteristics of the compounds. Rather than load the data into memory and scan through all bitmaps in the dataset, an N-dimensional GiST index (where N is the number of bits in the bitmap) can be generated that allows the rapid identification of just those structures with a high degree of similarity. Note that as the size (N) of the problem increases, the efficiency of the index-based approach (which is of order log N) increases when compared to that of the linear scan approach (which is of order N).

Another application area for GiSTXten is the analysis of medical data involving multiple groupings. For example, a large medical dataset could contain information on sex, age, weight, height, racial origin, medical procedures, drug prescriptions, history of illnesses, etc. A single GiST index would allow the rapid identification and extraction of just those records which matched multiple specified sets of predefined search criteria.

"By developing products that utilize the spatial enhancements in SQL Server 2008, our partners open up new doors for our joint customers," said Kim Saunders, senior director of SQL Server marketing at Microsoft Corp. "Barrodale offers a way for users to build custom GiST indexes that can provide faster searching."