??? Strict.time.constraints:.Especially true in NRT where the population cycle
can have a period of 10 or less minutes, but also in traditional batch systems
in particular circumstances.
??? Reliability/availability: Reliability is important, because in some contexts not
even one data item is allowed to be lost, which means that any missing data
must be recovered. The other face of reliability is availability, in our context,
the ability to recover from external/internal error (data loss or corrupted, etc.)
to maintain data integrity and consistency.
??? Loading.atomicity:.The system must be able to manage partial load in atomic
fashion (manage a sort of global commit) in order to allow partial recovery.
??? Source instability: In some contexts the sources of data are not stable (e.g.,
due to network problems or systems unavailability) and acquisition process
must be able to retry.
??? Nonstandard/complex transformation: DW techniques are not only applied
in ???sales and revenue??? contexts; in many cases one needs to implement
complex transformations or specialized calculations that differ from standard
analytical functions.
We have not yet mentioned issues like metadata and cleaning which are vastly
debated. Metadata are data that describe other data: for example, a column named
???price??? obviously refers to the price of an article but many other pieces of information
may also be useful for understanding its exact meaning.
Pages:
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192