The Evolution of Vertical Database Architectures - A Historical Review (Keynote Talk)

My intention in this lecture is to discuss the evolution of key concepts behind today’s emerging vertical database architectures. The Cantor project [5, 7] pioneered the analysis and coordinated application of many of these concepts in relational systems, which is one reason why references to this work are a recurring theme in what follows. The other reason is that although the work was duly reported in reasonably well-known conference publications, it has left no trace in citations. Thus, from a strictly evolutionary perspective, Cantor was a dead branch which left no progeny, but from a historical perspective it might still provide useful lessons. Transposed files as such were used in a number of early non-relational data base systems, mostly intended for statistical or scientific applications. A fairly comprehensive list of such systems was given by the paper [6] which is cited below. One great conceptual step that is now being taken is the realization that the adoption of transposed files opens a whole range of architectural opportunities. By careful combination of these opportunities dramatic performance gains may be provided, in particular of course when systems are used in those statistical and analytical kinds of application for which the concept was originally developed. So what do these architectural opportunities consist of? Below is a list, however, due to space limitations it is not possible here to give a fair account of all of them: