Overview

Can data be clustered different ways simultaneously without duplicating the data? There are times for all of us where we want to play it both ways, and times when we probably would like to play things much more than two ways. Live in the quiet of the country, or live in the action of the city. Enjoy the decadence of tasty foods, or enjoy the benefits of a healthy diet. Unfortunately, many of these situations are mutually exclusive. In the space of database design, data clustering long appeared to be one of these situations; a data set can be clustered (or grouped on disk) along a set of dimensions one way and one way only, unless the designer is willing to incur the overhead of duplicating the data for each unique clustering model. At least this has appeared to be the case for the past few decades before the introduction of multidimensional clustering (MDC) into industrial database systems. MDC enables a database system to cluster data along multiple dimensions at the same time. So, for example, data can be kept clustered by region of sale and by date of sale simultaneously. The essence of MDC is fundamentally to do the impossible: the clustering of data in different ways simultaneously without duplicating the data.

The author s first exposure to MDC was in the year 2000. About 15 people crammed into a small meeting...