To cope with this problem various indexing schemes
have been developed (Chan & Ioannidis, 1998; O??™Neil & Grafe, 1995; O??™Neil &
Quass, 1997; Sarawagi, 1997; Wu & Buchmann, 1998). Also precomputation of
aggregation results has been studied extensively??”mainly as a view maintenance
problem??”and is used as a means of accelerating query performance in data warehouses
(Roussopoulos, 1998; Srivastava, Dar, Jagadish & Levy, 1996).
However, for ad hoc star queries the usage of precomputed aggregation results is
extremely limited or even impossible in some cases. Even when elaborate indexes
are used, due to the arbitrary ordering of the fact table tuples, there might be as
many disk page accesses as are the tuples resulting from the fact table. The only
alternative one can have for such queries is a good physical clustering of the data,
and it is exactly for this reason that a new class of primary organizations for the
fact table has emerged (Karayannidis, Sellis, & Kouvaras, 2004; Markl, Ramsak,
& Bayern, 1999). These organizations exploit a special kind of key that is based on
the hierarchy paths of the dimensions, in order to achieve hierarchical clustering of
the facts. This physical clustering results in a reduced I/O cost for the majority of
star queries, which are based on the dimension hierarchies. Moreover, in a dimensional
data warehouse it is natural to exploit a multidimensional index for storing
the tuples.
Pages:
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277