STAR Model from OO/ODBMS View - 2
DST, ?DST do exist as well-defined entities in this scheme, as products of different processing stages
Algorithms associated with a data object are maintained with the object as associated operations, greatly easing management of consistency in an evolving and distributed software development effort
Data location management, including migration to and retrieval from the mass store, is done by the ODBMS via an interface to the mass store, HPSS
Other valuable capabilities, possibly candidates for ‘staging’ (can probably do without on day 1, and problems/limitations may motivate that, but design implications should be understood):
- Software evolution can be managed through ODBMS support for ‘schema evolution’, managing successive versions of objects and updating old objects to new formats
- Object versioning can support eg. ‘user’ versions of ‘public’ objects for development, private analysis
- Database replication can support live, updated database components at remote sites (important early on for online)