Information Technology Reference
In-Depth Information
model between the PSM and the hybrid PIM will record the
Derived from
traces,
whereas the horizontal trace model between the initial PIM and the hybrid
PIM will record the
Evolution
traces. An additional PIM2PIMTrace2 model can
be added to record the existing
Overlaps
and
Conflicts
, but these should be
manually added, since only the designer knows which elements in the model
refer to the same concept. It is important to note that, these kind of traces, are
not less important than the automatic ones, since they will act as a bridge to
map certain requirements to the data sources.
The last step is deriving the final PIM, which will be used to generate the
target DW. This final PIM retains only the elements from the hybrid PIM which
will be finally used. In this sense, this PIM is the result of filtering the undesired
elements and resolving the conflicts which appeared in the hybrid PIM. There-
fore, the traces from the hybrid PIM to the final PIM will show which concepts
were the ones chosen as a solution to each existing conflict. The type of these
traces will be
Evolution
and will be stored in the PIM2PIMTrace3 model.
Since the development process is performed by succesive deriving, adding,
and filtering elements while most elements are not altered, the traces have low
volatility. In this sense, developing a reactive framework which automatically
updates the corresponding traces whenever a change (update or delete) is made
would minimize the maintenance effort.
Once we have defined the trace metamodel and we have shown all the required
trace models, we need to formally define the automatic derivation of the traces
by means of QVT [20] rules.
4 Automatic Derivation of Traceability Models in Data
Warehouses
In this section, we will discuss the necessary transformations to automatically
generate the aforementioned traces and store them in trace models, which can be
updated over time. Due to paper constraints, we will focus on the traces coming
fromCIMtoPIM,bothinthissectionandinourexample.
According to our proposal for developing DWs [15, 16], we use a hybrid ap-
proach deriving the elements in an initial PIM model from the requirements
by means of QVT rules. QVT rules specify a transformation by checking for
a defined pattern in the source model. Once the pattern is found, a QVT rule
transforms elements from the source metamodel into the target metamodel. A
QVT, which creates the links between the CIM business process element and
the PIM fact and fact attribute elements, is shown in figure 5.
On the left hand of the transformation rule is the source metamodel, which
in our case, is our i* profile in order to model requirements for DWs. In this
QVT rule, we have a business process with its associated rationale, which is
modeled by means of strategic, decisional and information goals. These goals
model the business logic from which we obtain the information requirements
which, in turn are decomposed into measures, which are indicators of business
performance, and contexts.
