We note that this is a good base line since the projection
index is known for outperforming many multidimensional indices.
In the next set of experiments we measure the query size with the query box size. A
query box is a hypercube formed by the boundaries of the range conditions in the
attribute domains. We measure the query box size as the fraction of the query box
volume to the total volume of the attribute domains. If all attributes have uniform
distribution, then a query box size of 0.01 indicates that the query would select 1%
of the dataset. We say a query is more selective if the query box size is smaller.
Figure 9 shows the response time (in seconds) for 2- and 10-dimensional queries with
various query box sizes. For all experiments the query box size was chosen randomly
and covers the whole domain range. In general, we see that the query processing
time for the bitmap indices decreases as the queries become more selective. On the
Figure 8. Time (in seconds) to answer a one-dimensional range query using a WAH
compressed bitmap index is a linear function of the number of hits
4 Stockinger & Wu
Copyright ?© 2007, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of
Idea Group Inc. is prohibited.
other hand, the query processing time for the projection index stays constant as the
selectivity changes.
Pages:
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342