Queries with K1k (HC1 and HC4) show similar performance for both database
sizes. As the attribute cardinality decreases, REBSI performance improves.
For the PMap, performance for similar queries varies according to the position
of attribute properties in the pstring rather than attribute cardinalities, as
in the case of HCAQS (Gupta et al., 2002). Though KSEQ has cardinality 10
times that of K100k, the savings in the case of queries accessing K100k (HC3
and HC6) compared to similar queries accessing KSEQ (HC7 and HC8) are
greater. This is because of the relative positions of the properties covering these
attributes in the pstring. The difference between pfilterh and pfilterl is the main
contributing factor to the cost of index page retrieval in the worst case, i.e.,
PMax. The relative position of properties in the pstring has significant impact
on performance of a PMap.
3. Multi-attribute queries: For each high cardinality attribute KN, the PMap
retrieves fewer than or the same number of pages for queries of the form K2
AND KN as the ones with only KN. Multiple conditions increase the number
of bits that are set in the pfilterl, which reduces the difference between pfilterh
and pfilterl, resulting in fewer pages retrieved. For the REBSIs, a higher
number of attributes results in more pages retrieved as the number of scans
increases.
Pages:
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382