In the near future, it may be possible to see ETL tools/system written in
Java with sophisticated metadata support that works well in an application context
where today we have cryptic C programs. But, if it is true that machines run faster
and faster, even the volumes of data grow, so that there will always be the ???border
line??? applications where the performance constraints are again strong.
With regard to the possible spread of some kind of intersystems communication
standards/guidelines, we are a little more skeptical; these forms of standardization
do not have a strong economic boost, involve delicate internal organization balance,
and are really a too complex job. In our opinion, the ETL complexity is also
directly correlated to high volumes, timing constraints, reliability requirements,
Extraction, Transformation, and Loading Processes 0
Copyright ?© 2007, Idea Group Inc. Copying or distributing in print or electronic forms without written permission
of Idea Group Inc. is prohibited.
and so forth that need very sophisticated techniques. One of the main criticalities
in facing an ETL project consists in evaluating the impact of different choices in
order to weigh their costs in different perspectives (performance, etc.), and making
decisions respecting the balance of the system in its wholeness.
In this chapter, we have proposed an infrastructural approach to ETL as an optimal
solution for a specific class of problems in large DW; we have given some practical
suggestions in order to address typical implementation issues, leaving other aspects
and points of view in the background.
Pages:
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226