Databases Reference
In-Depth Information
Figure 8.2 Examples of three-dimensional MDC. Car sales by color, date of sale, and region: (a) all
data cells, (b) all sales of red cars, (c) all car sales in the province of Quebec, (d) sales of all red cars in
the province of Quebec, and (e) all cars sold during February 1999.
sorted by alphabetic rank as 1a, 3b, 2c, 5d, 7e, 6f, 4g. How does MDC make it possible
to cluster both numerically and alphabetically at the same time? By using two tricks.
First, by requiring high data density within the MDC cells. Consider how things would
change if we had a million 1a's and 1.5 million 2c's. With that kind of volume one could
put all the 1a's together in their own blocks, without worrying at all if the 1a's were any-
where near the 2c's on disk. Secondly, by not worrying too much about the physical
location of storage blocks on disk. The assumption with MDC is that provided the stor-
age blocks are large enough, the savings obtained from reading a single large block of
clustered data will achieve more benefit than the seek time caused by blocks not always
being adjacent on disk.
