The conceptual process perspective traces the reasons behind the structure of the data warehouse. We extend the demand-oriented concept of dependencies as in the Actor-Dependency model [68], with the supply-oriented notion of suitability that fits well with the redundancy found often in data warehouses. As an another extension to the Actor-Dependency model, we have generalized the notion of role in order to uniformly trace any person, program or data store participating in the system. By implementing the metamodel in an object logic, we can exploit the query facilities of the repository to provide the support for consistency checking of the design. The deductive capabilities of ConceptBase [28] provide the facilities to avoid assigning manually all the interdependencies of activity roles in the conceptual perspective. It is sufficient to impose rules to deduce these interdependencies from the structure of data stores and activities. While the design and implementation of the warehouse are performed in a rather controlled environment, the administration of the warehouse has to deal with problems that evolve in an ad-hoc fashion. For example, during the loading of the warehouse contingency treatment is necessary for the efficient administration of failures. In such events, not only the knowledge of the structure of a process is important; the specific traces of executed processes are also required to be tracked down in an erroneous situation, not only the causes of the failure, but also the progress of the loading process by the time of the failure must be detected, in order to efficiently resume its operation. Still, failures during the warehouse loading are only the tip of the iceberg as far as problems in a data warehouse environment are concerned. This brings up the discussion on data warehouse quality and the ability of a metadata repository to trace it in an expressive and usable fashion. To face this problem, the proposed process metamodel is explicitly linked to our earlier quality metamodel [32]. We complement this linkage by mentioning specific quality factors for the quality dimensions of the ISO 9126 standard for software implementation and evaluation. Identifying erroneous situations or unsatisfactory quality in the data warehouse environment is not sufficient. The data warehouse stakeholders should be supported in their efforts to react against these phenomena. The above-mentioned suitability notion in the conceptual perspective of the process metamodel allows the definition of recovery actions to potential errors or problems (e.g., alternative paths for the population of the data warehouse) in a straightforward way, during runtime. Data warehouse evolution is unavoidable as new sources and clients are integrated, business rules change and user requests multiply. The effect of evolving the structure of the warehouse can be predicted by tracing the various interdependencies among the components of the warehouse. We have already mentioned how the conceptual perspective of the metamodel traces interdependencies between all the participants in a data warehouse environment, whether persons, programs or data stores. The prediction of potential impacts (whether of political, structural, or operational nature) is supported by this feature in several ways. To mention the simplest, the sheer existence of dependency links forecasts
a potential impact in the architecture of the warehouse in the presence of any changes. More elaborate techniques will also be provided in this paper, by taking into account the particular attributes that participate in these interdependencies and the SQL definitions of the involved processes and data stores. Naturally, the existence of suitability links suggests alternatives for the new structure of the warehouse. We do not claim that our approach is suitable for any kind of process, but focus our attention to the internals of data warehouse systems. This paper is organized as follows: In Section 2 we present the background work and the motivation for this paper. In Section 3 we describe the process metamodel and in section 4 we present its linkage to the quality model. In section 5 we present how the metadata repository can be used for the determination of the operational semantics of the data warehouse tables and for evolution purposes. In section 6 we present related work and section 7 presents issues for future research.