Algebraic Graph-based Approach to Schema Integration

The language of sketches is as concise and powerful as higher order logic and as handy as conventional ER-diagrams. We discuss its beneets in respect to problems of schema integration. It is evident that a chief property of the next generation information systems is their organization on cooperative principles. Typically, applications involved in a cooperative system have to access data stored in several separate databases. These databases can be diierently organized, semantically overlapping and even inconsistent , so that an application programmer encounters severe problems, moreover, in each multibase case they have to be solved speciically. In fact, as it was noted in 13], the situation is similar to the multiile situation before invention of DBs, and the analogous solution can be suggested: in order to provide applications with a single integral view of data, local DBs should be integrated into a united distributed DB. Under the latter we mean a database system which has a schema like an ordinary DB but its extension is virtual, that is not stored immediately but derived from the extensions of local databases. It is characteristic to multibase systems that local DBs can be autonomous, oriented on applications of diierent level, and independently developed. This results in the essential heterogeneity of whole systems. Thus, heterogeneous multibase integration turns out to be a fundamental issue of cooperative system operation. In particular, in the context of federated database environment, integration is a function regularly performed at diierent levels and by diierent services depending on the organization of the federal environment. The database integration gives rise to several problems: Heterogeneous framework: If data are stored in diierent systems, the database schemas can be deened in diierent languages; we need a way to compare such deenitions. Structural connicts: Semantically the same data can be presented by diierent constructs (function or relation; attribute or entity). Note, these are connicts between presentations, not between data.