Centralized DB systems

Distributed DB systems

the data is stored at several sites
each site is managed by a DBMS that can run independently; these autonomous components can also be heterogeneous

⇒ impact on: query processing, query optimization, concurrency control, recovery

gateway protocols

distributed data independence (extension of the physical and logical data independence principles):
- users can write queries without knowing / specifying the actual location of the data
- cost-based query optimization that takes into account communication costs & differences in computation costs across sites
distributed transaction atomicity
- users can write transactions accessing multiple sites just as they would write local transactions
- transactions are still atomic (if the transaction commits, all its changes persist; if it aborts, none of its changes persist)