Computational database structures
D. Stott Parker, Hung‐Chih Yang · 2006
Emerging software systems in disciplines like eScience process vast amounts of diverse data at amazing speed and in a computationally-intensive way. To better serve these systems, integrating data management and computational services is a better way to enhance DBMS on performance and functionalities than separating or aggregating them. Computational database structures , such as the Holodex, Lightweight Functions, and SB/HiST, emphasize integration at the structural level, thus they still can achieve separation of concerns at the architectural level. The Holodex is a general-purpose integration of an index and hierarchy-oriented programming constructs. It can efficiently compute some nonparametric statistics. Lightweight Functions integrates programming logic (as functions) with the relational data model, so that functions can be managed like data. Databases can be designed and implemented to follow the table-driven programming methodology. The SB/HiST is an integration of histograms and a tree index structure. Derived density estimates are great resources of data characteristics that can be used for optimizing queries and guiding index traversals. If armed with a proper density estimate, the SB/HiST can outperform the B+-Tree because of its maximal fan-out.