On the
contrary, if DW loading frequency, volumes, and complexity are high, then a more
structured approach becomes necessary.
In these cases it is necessary to find out sophisticated ETL techniques, whereas in
other cases the more usual approach is the best solution. It is true, however, that
if a company has bought licenses for a commercial tool or has developed its own
infrastructure (as we propose here), then it is strongly recommended to use them
in all kinds of DW projects.
In a context of strong requirements, we can individuate the most common constraints
and criticalities that one can meet in developing an ETL system:
flat file
source dbms
transformations
data validations
data integration
aggregation
loading
target dbms
ETL
dbms/system/application
Operation&Maintenance
other sources
other calculations
cleaning
Extraction/acquisition
Figure 1. ETL scenario
Extraction, Transformation, and Loading Processes
Copyright ?© 2007, Idea Group Inc. Copying or distributing in print or electronic forms without written permission
of Idea Group Inc. is prohibited.
??? Volumes: High volumes imply specialized loading techniques (classic SQL
is not suited) and require a good efficiency of transformation code.
??? Near real time (NRT): In these contexts, loading and analysis processes must
coexist together, and that implies strong constraints on loading techniques;
massive loading mechanisms are not directly applicable, indexes must be active,
and so on.
Pages:
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191